Issue a rotatable external-only TCP token on enable, soft-revoke it on disable without killing long-lived stdio sockets, force external sockets onto the reserved chat scope, and only seed empty external metadata when the scope key was never written.
Co-authored-by: Cursor <cursoragent@cursor.com>
Reject disabled/in-flight scope resurrection, move session sync to App mount, cancel the external chat session on stop, and broadcast approval cards to main + settings windows.
Co-authored-by: Cursor <cursoragent@cursor.com>
Keep app-wide external hosts when a Catty sidebar push arrives, and re-read discovery permissionMode so long-lived stdio clients pick up Safety changes without restart.
Co-authored-by: Cursor <cursoragent@cursor.com>
In terminal-worker mode mirror renderer session payloads into
__external_mcp__ and seed from other chat scopes when the live map is empty.
Co-authored-by: Cursor <cursoragent@cursor.com>
In terminal-worker mode the main-process sessions map can be empty; avoid
overwriting renderer-pushed __external_mcp__ metadata with an empty sync.
Co-authored-by: Cursor <cursoragent@cursor.com>
Merge prior scoped host metadata when syncing live sessions, and parse
Claude Environment: KEY=VALUE blocks so configured entries are not
misclassified as missing discovery env.
Co-authored-by: Cursor <cursoragent@cursor.com>
Disconnect marked external sockets on disable, await in-flight stop before
re-enable, write Claude MCP entries to user scope, and stop spawning client
CLIs on the quiet status poll.
Co-authored-by: Cursor <cursoragent@cursor.com>
Expose the existing catalog MCP as a Settings-managed local bridge with
discovery lifecycle, one-click client setup, and reserved external scope
so third-party agents can use Netcatty without a sidebar chat session.
Co-authored-by: Cursor <cursoragent@cursor.com>
Clarify Netcatty MCP environment context so external agents that read resources can see the terminal execution tools, and cover the MCP resource path with a test.
* Run agent commands in a subshell so failures cannot close the session
Agent-generated commands were eval'd directly in the user's active login
shell, so a generated script containing `set -e` (or `exit`) terminated
the whole SSH session on the first failure. Wrap the eval in a subshell
unconditionally instead of trying to detect dangerous commands — the
detection approaches from PRs #1852/#1882 grew into shell parsing and
were abandoned. Output stays live in the user's terminal; the trade-off
is that cd/export no longer persist between calls, which the tool
descriptions now tell the model about.
Also fix multi-line command display: synthetic echo now normalizes line
breaks to CRLF (no more staircase indentation in xterm), and the POSIX
wrapper rebuilds newlines via printf so the typed command stays on one
physical line and PS2 continuation echoes ("> ...") no longer leak past
the marker filter.
Fixes#1850
Co-authored-by: Cursor <cursoragent@cursor.com>
* Move shellUtils require to module scope in cattyExecHandlers
The inline require("./ai/shellUtils.cjs") inside the with(ctx) block did
resolve correctly at runtime (bare `require` hits ctx.require, based in
electron/bridges/), but the path only works through that indirection and
reads as wrong relative to this file. Require at module scope with the
unambiguous ../ai path instead.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
* feat(ai): unify capability registry, harness tools, and agent runtime
Centralize MCP/CLI/Catty tool surfaces in the capability catalog with codegen,
move Catty-only harness tools into catalog with local execution, converge
mcpServerBridge builtin dispatch via registry, and wire AgentRuntime with
vault bridge, permission grants, and Round 2 catalog parity.
Co-authored-by: Cursor <cursoragent@cursor.com>
* feat(ai): vault host tools, artifact cards, and external agent vault guidance
Unify vault host create/import with capability RPC routing, add chat artifact cards with Vault navigation, and align external MCP agents with Catty rules so host creation never falls back to Vault Notes.
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix(ai): address Bugbot findings for streaming, vault bridge, and grants
Clear Catty streaming by chat sessionId, read latest vault state via getters on agent mutations, and stack grant persisters so unmounting one AI panel does not break others.
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix(vault): clear pending openNoteId after chat navigation focus
NotesManager notifies VaultView once the one-shot openNoteId request is applied so the id does not stick for the rest of the session.
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix(vault): serialize vault agent IPC mutations to prevent lost writes
Queue vault agent bridge handlers so concurrent hosts.create or note.create tool calls cannot overwrite each other via stale read-modify-write snapshots.
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix(vault): keep synchronous vault snapshot for agent IPC reads
Mirror hosts, notes, and groups in a ref updated immediately on agent writes so serialized mutations see the latest state before React re-renders.
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix(ai): sync Catty MCP sessions and defer runtime cleanup on delete
Register terminal sessions with the main process before Catty tool calls, and clear ToolOutputStore only after stopping an active turn when deleting a chat session.
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix(ai): avoid deleting a newer abort controller on late stop
Only remove the session abort-controller entry when stop finishes if it still refers to the controller that was stopped.
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix(ai): sync vault snapshot on render and await active turn on stop
Refresh vault agent reads when React props change without clobbering in-flight agent writes, and wait for runTurn to finish before clearing ToolOutputStore on session delete.
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix(ai): reload permission grants from storage before auto-approve
Re-read persisted grant rules on each approval check so revocations in Settings take effect without requiring a chat hook remount.
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix(ai): sync permission grants to main process immediately on write
Push MCP permissionGrantsSnapshot as soon as Always allow persists a rule so the next tool call in the same agent turn sees the grant without waiting for useEffect.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
* feat(capabilities): add shared agent capability layer for MCP and CLI
Introduce a registry-driven capability foundation so terminal, SFTP, vault,
and port-forward tools can be extended from one policy and surface definition.
Wire mcpServerBridge permission checks and the Skills CLI RPC client through
the shared layer without changing existing AI runtime behavior.
Co-authored-by: Cursor <cursoragent@cursor.com>
* fix(capabilities): restore SFTP cancel gating and home RPC timeout
Match legacy mcpServerBridge behavior by blocking SFTP write RPCs after
chat cancel and treating sftp/home as a long-running operation for timeout
budget calculation.
Co-authored-by: Cursor <cursoragent@cursor.com>
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
Follow-up to #851 (Codex review comment on 32bab2d4). After that PR,
`resolveEffectiveShellKind` flips an unknown-shell session to PowerShell
based on `session.lastIdlePrompt`, but that field is updated only when
`trackSessionIdlePrompt` recognizes a known prompt shape (default
PowerShell or `user@host[:path][#$]`). On an SSH/Telnet session that
enters PowerShell and then leaves it for a shell with an unrecognized
prompt — cmd.exe (`C:\>`), oh-my-posh / starship / a custom PS1 — the
cached `PS ...>` value persists indefinitely, and every subsequent MCP
command keeps getting wrapped as PowerShell against a non-PowerShell
shell. The new shell errors on the wrapper syntax once per command, and
nothing self-heals until the user reconnects.
Add `getFreshIdlePrompt(session)` which returns the cached prompt only
when the rolling PTY tail (`session._promptTrackTail`) still ends with
it. If the visible last line has moved on — even to a prompt shape we
don't recognize — the cache is treated as expired and downstream
wrapper selection / suffix matching falls back to `shellKind` alone,
which is the correct behavior for the unknown-shell case.
Wire this into the three call sites that previously read
`session.lastIdlePrompt || ""`:
- `aiBridge.cjs:1325` (Catty Agent foreground exec)
- `mcpServerBridge.cjs:1496` (MCP `terminal_execute`)
- `mcpServerBridge.cjs:1584` (MCP `terminal_start` background job)
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Add background terminal jobs for long AI commands
* Bound background job output buffering
* Fix long-running terminal job polling and stop behavior
* Fix terminal job final output and stopping retention
* Wait for PTY stop confirmation before cancelling
* fix: address codex review findings in PTY job refactor
- [P1] Use last occurrence of start marker to skip echoed wrapper command,
preventing control markers from leaking into stdout
- [P1] Add wall-clock timeout for foreground PTY execution so commands that
print continuously still get terminated at the configured limit
- [P2] Add hard deadline for cancellation so jobs that ignore Ctrl+C are
force-finished after 30s instead of staying stuck in "stopping" forever
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-2 codex review findings
- [P1] Use visibleOutput for background job completion to keep offsets
consistent with polling, preventing output loss when raw buffer
(with ANSI codes) truncates earlier than the visible buffer
- [P2] Clarify system prompt that terminal_start requires PTY-backed
sessions, so exec-only SSH sessions are not incorrectly routed
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-3 codex review findings
- [P1] Always strip markers from visibleOutput in background job finish
to prevent end-marker lines leaking into terminal_poll results
- [P2] Correct terminal_execute timeout guidance from ~2min to ~60s to
match the actual default commandTimeoutMs (60000)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-4 codex review findings
- [P1] Delay session lock release when cancel is forced (process may
still be running) to prevent sending commands into a busy shell
- [P2] Move scope validation before pendingSessionWriteApprovals so
out-of-scope requests fail fast without blocking the write lock
- [P2] Add session scope checks to handleJobPoll and handleJobStop
so chats that lose access cannot read output or cancel jobs
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-5 codex review findings
- [P1] Strip marker lines before they enter the bounded visible buffer
so they never occupy space or leak as partial fragments on truncation
- [P2] Never release session lock after forced cancellation since the
previous process may still be attached to the PTY
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-6 codex review findings
- [P2] Buffer incomplete marker lines across PTY chunks to prevent
partial marker fragments from leaking into visible output
- [P1] Release session lock after 60s delay on forced cancel as
compromise between safety and permanent lock
- [P2] Enforce session scope checks on jobPoll/jobStop for both
dynamic (chatSessionId) and static (NETCATTY_MCP_SESSION_IDS) modes
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-7 codex review findings
- [P2] validateSessionScope now accepts explicit scopedSessionIds so
static MCP scope mode is enforced for jobPoll/jobStop too
- [P2] Apply per-session execution lock to netcatty:ai:exec IPC path
so it cannot race with active background jobs on the same session
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-8 codex review findings
- [P1] Make wall-clock timeout opt-in via enforceWallTimeout flag,
enabled only for MCP terminal_execute path. Catty Agent's
netcatty:ai:exec keeps the inactivity-based timeout since it has
no terminal_start fallback for long-running streaming commands
- [P2] Always allow handleJobStop regardless of session scope so
the per-session execution lock can always be released after
workspace membership changes
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-9 codex review findings
- [P1] Enable enforceWallTimeout for netcatty:ai:exec to match the
pre-PR behavior (hard wall-clock deadline). Without this, tail -f
or verbose builds would hold the session lock indefinitely
- [P2] Treat explicit scopedSessionIds=[] as no access rather than
falling through to global scope, matching handleGetContext's
documented behavior
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-10 codex review findings
- [P2] Add bounded startup deadline (30s) for the start marker arrival
even when wall-clock timeout is disabled. Prevents background jobs
from hanging indefinitely on already-chatty PTY sessions
- [P3] Use job-specific marker (not generic __NCMCP_) when stripping
marker lines, so user output containing __NCMCP_ is preserved
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-11 codex review findings
- [P2] Skip the 30s startup timeout for foreground execViaPty paths.
It now applies only when maxBufferedChars > 0 (background jobs),
so foreground commands queued behind a busy shell can wait
- [P2] Return empty stdout from getSnapshot() before the start marker
arrives, so an early poll cannot advance nextOffset past pre-start
PTY noise that gets discarded once the real command begins
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-12 codex review findings
- [P1] Treat empty chat scopes as no access in validateSessionScope:
if a chat has explicit scoped metadata (even []), enforce strictly
rather than falling through to fallback/global scope
- [P2] Re-add session scope check in handleJobStop for static MCP
clients (scopedSessionIds), while still allowing dynamic chat-scoped
callers to always stop their own jobs even after scope changes
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-13 codex review findings
- [P2] getScopedJob now requires the caller to present the job's
chatSessionId. Unscoped/static callers cannot reach into another
chat's background jobs even if they learn the jobId
- [P2] Stop button no longer cancels terminal_start background jobs.
They are intentionally long-running, so killing them on every
per-response stop defeats the purpose of the feature. Cleanup on
chat deletion (cleanupScopedMetadata) is preserved
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-14 codex review findings
- [P1] terminal_start jobs no longer registered in activePtyExecs so
ACP "Stop" / cancelPtyExecsForSession does not kill them. They are
still managed via terminal_stop and the per-session execution lock
- [P1] Remove enforceWallTimeout from netcatty:ai:exec since Catty
Agent has no terminal_start fallback for long-running commands.
Inactivity timeout still catches genuinely hung processes
- [P2] Forced-cancelled jobs stay in "stopping" (completed=false)
until the 60s lock grace period ends, so callers don't see the
job as completed while the session is still locked
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-15 codex review findings
- [P2] Allow netcatty/jobStop to bypass the chat-cancelled gate so
users can stop terminal_start jobs even after ACP "Stop" was pressed
- [P2] Mark non-zero exit codes as failed (not completed) so callers
don't have to special-case exitCode against status
- [P2] Pre-start cancel: clear startup timer in requestCancel and
detect prompt return on preStartOutput so a queued job that gets
cancelled resolves as "Cancelled", not "startup timed out"
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-16 codex review findings
- [P2] Cap preStartOutput for background jobs at maxBufferedChars so
noisy idle PTYs cannot accumulate megabytes before the start marker
arrives or the startup timeout fires
- [P2] On forced cancel, immediately release the session lock and
mark the job as cancelled. The error message clearly states that
the process may still be running, and the caller sees completed=true
exactly when the lock is no longer held — consistent semantics
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-17 codex review findings
- [P2] Disable prompt-suffix completion fallback for background jobs.
Long-running commands often print prompt-like text (nested shells,
ssh, sudo -s, REPLs) and would otherwise be misdetected as completed.
Background jobs rely strictly on the end marker
- [P2] consumeVisibleText now treats \\r as a carriage return that
resets the current line, so progress bars (npm, docker pull, curl)
collapse to the latest frame instead of accumulating every redraw
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-18 codex review findings
- [P2] Pre-start cancel on sessions without a tracked idle prompt now
gets a 2s fallback to finish as Cancelled, instead of waiting the
full forced-cancel window for an end marker that will never arrive
- [P3] Move session-scope validation before the busy-session check so
out-of-scope callers cannot probe the existence/activity of foreign
sessions via busy-state error messages
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-19 codex review findings
- [P1] Re-enable prompt-suffix completion fallback for background
jobs but with a longer 10s delay so nested shells / REPLs have
time to print past their initial prompt before the recheck
- [P2] Carriage returns now collapse progress redraws across PTY
chunks: \\r is preserved through consumeVisibleText and
applyCarriageReturns erases the trailing line of visibleOutput
when a chunk starts with \\r. Verified with a fake PTY that
emits "10%" then "\\r20%" then "\\r30%\\n" — final output is "30%"
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-20 codex review findings
- [P1] Disable prompt-suffix completion fallback for background jobs.
Commands that open child shells with the same prompt as the parent
(bash, zsh, sudo -s, ssh) would otherwise be reported as completed
while the child is still running. Background jobs rely strictly on
the end marker, with their long timeout and explicit terminal_stop
- [P2] Track a monotonic visibleHighWatermark so polling nextOffset
cannot move backwards across CR redraws. serializeBackgroundJob now
returns the latest visible frame when the caller's offset has been
passed by a redraw, instead of returning empty stdout permanently
- [P3] Buffer trailing lines that contain the constant __NCMCP_
prefix (not just the full random marker token) so PTY chunk
boundaries that split the marker mid-token cannot leak _E:0 noise
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-21 codex review findings
- [P2] Foreground execs now also get a hard startup deadline (using
the configured timeoutMs as the limit). Background jobs use a
fixed 30s. Without this, an already-chatty PTY would let onData
re-arm the inactivity timer forever before _S arrives
- [P2] finish() now uses the monotonic visibleHighWatermark for
totalOutputChars on completion, so the final poll's nextOffset
cannot regress relative to earlier polls after CR redraws
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-22 codex review findings
- [P2] cleanupScopedMetadata now also calls clearPendingApprovals so
in-flight approval requests resolve immediately. Otherwise a chat
deleted while an approval was pending would leave the per-session
write lock held until the 5-minute approval timeout expires
- [P2] Allow netcatty/jobStop in observer mode so users can stop
long-running terminal_start jobs that were launched before they
switched to observer mode
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-23 codex review finding
- [P2] Apply \\r as a "deferred" carriage return: park the cursor at
the start of the line but defer erasure until the next character
arrives. This preserves the latest visible frame for commands like
printf '10%%\\r'; sleep; printf '20%%\\r' that pause between
redraws, while still collapsing continuous progress redraws to a
single frame. Verified: snapshots now show '40%' and '50%' instead
of empty stdout
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-24 codex review findings
- [P1] Re-enable prompt fallback for background jobs with a 30s
delay so commands open child shells / REPLs have time to print
past their initial prompt before the recheck. This is the third
time codex has flip-flopped on this — 30s is the compromise
- [P2] Pass chatSessionId to execViaChannel in handleExec so
cancelPtyExecsForSession can interrupt SSH exec-channel commands
scoped to the originating chat
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-25 codex review finding
- [P1] Stop in-place CR collapsing in visibleOutput. The collapsed
buffer made polling offsets non-monotonic and could drop finalized
lines after a CR rewrite. Now visibleOutput stores raw bytes (with
\\r dropped at consumeVisibleText to keep the buffer simple), the
256KB cap naturally bounds progress-bar accumulation, and slice
semantics work correctly across all redraw patterns. Consumers
that want a "collapsed view" can post-process
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: address round-26 codex review findings
- [P2] Carriage returns are now preserved in the raw buffer and
collapsed at serialize time in collapseCarriageReturns. This keeps
monotonic offsets in the buffer while polled output shows the
latest progress frame. A trailing \\r leaves existing content
intact (deferred erasure semantics)
- [P2] netcatty/jobStop now bypasses the confirm-mode approval gate
so a runaway terminal_start job can always be interrupted, even
when the renderer is unavailable
- [P3] requestCancel's one-shot timers (2s pre-start, 150ms reinforce,
30s force-finish) are now tracked and cleared in finish() so they
cannot keep the Node event loop alive after the job has resolved
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: bincxz <16399091+binaricat@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add COPILOT_HOME cleanup in list-models finally block to prevent
temp directory accumulation on each model fetch
- Remove verbose console.log in mcpServerBridge dispatch/connect/auth
that fired on every MCP call for all agents
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Adapt copilot agent additions to the refactored managed agent
architecture (resolveAgentPath + buildManagedAgentState pattern).
Add copilot to ManagedAgentKey type and MANAGED_AGENT_META.
Keep main's resolveMcpServerRuntimeCommand (process.execPath +
ELECTRON_RUN_AS_NODE) over PR's runtimeCommand parameter approach.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat: add deviceType field to Host model for network device support
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat: pass deviceType through session metadata pipeline
Thread deviceType from Host model through AITerminalSessionInfo, IPC
types, and mcpServerBridge so AI agents can inspect device type per session.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat: route network device SSH sessions to raw PTY execution
When deviceType === 'network', handleExec now uses execViaRawPty
instead of execViaPty so vendor CLIs (Huawei VRP, Cisco IOS, etc.)
receive commands as-is without POSIX shell wrapping or markers.
The command blocklist is also skipped for network devices, consistent
with the existing serial session bypass. AI context description updated
to document the raw-execution behaviour for network device sessions.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat: add network device mode toggle to host settings UI
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat: add network device awareness to Catty Agent system prompt
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: extend network device mode to Catty Agent exec path and host context
- Add network device detection and raw execution routing to aiBridge.cjs
(the primary Catty Agent command path), not just the MCP bridge
- Export getSessionMeta from mcpServerBridge for reuse in aiBridge
- Surface deviceType in Catty Agent system prompt host list so the AI
can identify which sessions are network devices
- Pass deviceType through buildSystemPrompt context
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: exempt network device sessions from client-side blocklist and update ACP context
- Add deviceType to ExecutorContext sessions type
- Skip renderer-side command blocklist for deviceType=network sessions
in shared toolExecutors.ts (not just main-process side)
- Update ACP agent context hint to mention network device sessions
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: only show network device mode toggle for SSH hosts
Telnet and local hosts don't support the network device execution path,
so hiding the toggle prevents users from enabling a broken configuration.
Serial hosts already use raw mode by default.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: exclude Mosh sessions from network device raw execution path
Mosh uses a shell-backed PTY and cannot connect to vendor CLIs, so
network device mode should only apply to SSH and serial sessions.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: prefer session.protocol over metadata for Mosh detection
Mosh tabs report protocol:"ssh" in renderer metadata but "mosh" in
the main-process session object. Prioritize session.protocol (runtime
truth) to correctly exclude Mosh from network device raw execution.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: suppress deviceType metadata for Mosh sessions
Mosh requires a shell-backed PTY and cannot connect to vendor CLIs,
so omit deviceType from AI-facing metadata when session is Mosh-backed.
This prevents the AI from being told to use vendor CLI syntax when the
actual execution path uses normal shell wrapping.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: use exit code 0 for network device sessions and hide toggle for Mosh
- Network device / serial sessions return exitCode: null from vendor
CLIs. Default to 0 instead of -1 so the AI doesn't misinterpret
successful commands as failures.
- Hide the network device mode toggle when Mosh is enabled, since
the setting is suppressed at runtime for Mosh sessions anyway.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: preserve null exit codes and restrict raw mode to SSH/serial
- Preserve exitCode: null for network device sessions instead of
coercing to 0, so the AI knows exit status is unavailable rather
than seeing a misleading success code.
- Explicitly whitelist SSH/serial protocols for network device mode
instead of just excluding mosh, preventing local/telnet sessions
from accidentally entering raw execution.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: use UTF-8 encoding for SSH network device raw execution
execViaRawPty hardcodes latin1 for serial port data decoding. Add an
encoding option (default: latin1) and pass utf8 from SSH network
device call sites so multi-byte characters aren't corrupted.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: use host charset for serial port decoding instead of hardcoded latin1
- Extract charsetToNodeEncoding() to module scope in terminalBridge
- Serial sessions now read options.charset (from Host.charset) for
both terminal display decoding and AI command output
- Store serialEncoding on session object so exec paths can use it
- Pass encoding through all execViaRawPty call sites
- Default encoding changed from latin1 to utf8 (matches most modern
network equipment and is the safer default for CJK environments)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: move serialEncoding declaration before session object creation
serialEncoding was referenced in the session object literal before its
const declaration, causing a TDZ ReferenceError that would crash every
serial connection.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: tighten isNetworkDevice logic and clean up edge cases
- Align toolExecutors isNetworkDevice check with bridge logic: require
explicit SSH/serial protocol match instead of trusting deviceType alone
- Remove empty-string protocol match from isSshOrSerial in both bridges
to prevent local/unknown sessions from being treated as network devices
- Widen exitCode return type to `number | null` to match actual behavior
- Clear deviceType when enabling Mosh (incompatible combination)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: update MCP server tool descriptions for network device sessions
The get_environment and terminal_execute tool descriptions only
mentioned serial/raw sessions for network devices. Updated to also
reference deviceType: network SSH sessions so external AI agents
(Claude, Codex) know about the new execution mode.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: include deviceType in get_environment and guard execViaChannel fallback
- Add deviceType to executeWorkspaceGetInfo session mapping and return
type so Catty Agent's get_environment tool matches MCP bridge output
- Guard both aiBridge and mcpServerBridge against falling through to
execViaChannel for network device sessions — network devices require
an interactive PTY and exec channels would produce broken behavior
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat: add charset setting to serial host configuration UI
Serial hosts now have a charset input in the Advanced section,
defaulting to UTF-8. The value is saved to Host.charset and used
by the serial decoder in terminalBridge.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat: add charset to serial quick-connect modal with full pipeline
- Add charset input to SerialConnectModal (Advanced section)
- Thread charset through onConnect callback → handleConnectSerial →
createSerialSession → TerminalSession.charset
- Add charset field to TerminalSession interface
- Include charset in fallback host builder for quick-connect sessions
so createTerminalSessionStarters can pass it to startSerialSession
- Saved hosts also store charset via onSaveHost
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: constrain serial connect modal height with scrollable content
Modal content could overflow the viewport when Advanced section was
expanded. Add max-h-[85vh] to DialogContent with flex layout so the
content area scrolls while header and footer buttons stay visible.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: propagate charset through all serial session creation paths
- Add charset to startSerialSession type in global.d.ts
- Copy host.charset to TerminalSession in connectToHost serial path
- Copy host.charset in createWorkspaceWithHosts serial path
- Propagate session.charset in splitSession (both workspace and standalone)
- Propagate session.charset in copySession
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* fix: propagate charset in remaining session creation paths
Add host.charset to connectToHost (non-serial), createWorkspaceWithHosts
(non-serial), and runSnippet session creation for consistency.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat: inline approval gate for tool execution
Replace SDK-level needsApproval with Promise-based approval gate inside
tool execute functions. The SDK stream stays alive while the UI shows
inline approve/reject buttons on ToolCall blocks.
Changes:
- Add approvalGate.ts: Promise-based approval system with event listeners
- tools.ts: requestApproval() inside execute for confirm mode
- tool-call.tsx: inline approval buttons and keyboard shortcuts
- ChatMessageList.tsx: subscribe to approval events, render approval UI
- useAIChatStreaming.ts: remove old useToolApproval hook integration
- AIChatSidePanel.tsx: remove old approval hook, clean up unused destructuring
- systemPrompt.ts: update confirm mode to not ask for text confirmation
- preload.cjs: filter pager env var prefixes from terminal display
- mcpServerBridge.cjs: add approval gate for ACP/MCP write operations
- aiBridge.cjs: wire IPC for MCP approval response and main window getter
- preload.cjs: add onMcpApprovalRequest/respondMcpApproval APIs
* fix: scope approval gate by chatSessionId and replay for late subscribers
Address Codex PR review comments:
- Add chatSessionId to ApprovalRequest for session isolation
- Scope clearAllPendingApprovals(chatSessionId?) to only clear
approvals belonging to the target session
- Add replayPendingApprovals() so late-mounting ChatMessageList
picks up approvals that fired while unmounted
- Scope MCP clearPendingApprovals in aiBridge cancel handler to
effectiveChatSessionId instead of clearing all
- Pass chatSessionId through MCP approval IPC flow
* chore: remove old approval flow code
- Delete useToolApproval.ts (unused hook)
- Delete InlineApprovalCard.tsx (replaced by ToolCall inline buttons)
- Remove stale comments referencing old hook in AIChatSidePanel
- Remove unused ai.chat.toolApprovalTitle i18n key from en/zh-CN
* fix: session-scoped approval gate and MCP replay survival
- handleStop passes activeSessionId to clearAllPendingApprovals
- setupMcpApprovalBridge stores MCP approvals in pendingApprovals map
so they survive ChatMessageList unmount/remount cycles
- ChatMessageList accepts activeSessionId prop and filters standalone
MCP approval blocks to the current session only
- AIChatSidePanel passes activeSessionId to ChatMessageList
* fix: filter PTY exec marker echoes and exit code lines from terminal
Extend filterMcpMarkers in preload.cjs to strip all shell-visible
artifacts from AI command execution:
- Echoed printf start marker: printf '%s\n' '__NCMCP_..._S'
- Echoed exit code restoration: (exit $__nc)
- PowerShell: Write-Output, $global:LASTEXITCODE, $__nc assignment
- Fish: set __nc $status
- Cmd: echo __NCMCP_...
- Widen guard to also trigger on __nc and PAGER=cat strings
* fix: scope SDK approvals, deny MCP on no renderer, fix memo comparator
- createCattyTools accepts chatSessionId and passes it to
requestApproval so SDK approvals can be matched by
clearAllPendingApprovals(activeSessionId) on stop
- useAIChatStreaming passes sessionId to createCattyTools
- mcpServerBridge: deny (resolve false) when no renderer window is
available instead of auto-approving, preserving confirm mode safety
- ChatMessageList: add activeSessionId to React.memo comparator so
switching sessions triggers re-render for correct MCP approval filter
* fix: MCP listener lifecycle, approval timeout, and UI sync on stop
- Move setupMcpApprovalBridge from ChatMessageList to AIChatSidePanel
so the IPC listener survives tab/panel switches
- Add 5-minute auto-deny timeout to requestApproval to prevent
indefinite isStreaming hangs when user walks away
- Add onApprovalCleared listener system: clearAllPendingApprovals now
notifies UI subscribers so ChatMessageList removes stale cards
- ChatMessageList subscribes to onApprovalCleared to sync local state
* fix: main-process approval timeout and full tool args in payload
- Add 5-minute auto-deny timeout to requestApprovalFromRenderer
matching the renderer-side requestApproval behavior
- Forward all tool params (excluding chatSessionId) to approval UI
instead of cherry-picking command/input/path, so sftpRename
oldPath/newPath and other tool-specific args are visible
* fix: move MCP bridge to TerminalLayer, narrow terminal filter guard
- Move setupMcpApprovalBridge from AIChatSidePanel to TerminalLayer
so the IPC listener stays alive regardless of side panel tab.
AIChatSidePanel only mounts when activeSidePanelTab==='ai'.
- Narrow preload.cjs filter guard back to __NCMCP_ only, preventing
false-positive stripping of user scripts containing __nc or PAGER=cat
* fix: eliminate PTY wrapper echo leakage and duplicate prompts
- Posix wrapper now emits 2 lines instead of 4: start marker + command
on line 1 (joined with ;), end marker + exit on line 2. This
eliminates the duplicate prompt echo from the separate start marker.
- Rename __nc to __NCMCP_rc in all shell variants (posix/fish/powershell)
so every wrapper variable contains the __NCMCP_ prefix. The preload
guard `data.includes("__NCMCP_")` now reliably catches ALL wrapper
artifacts regardless of chunk boundaries.
- Update all filterMcpMarkers regex patterns to match the restructured
wrapper format and renamed variable.
* fix: sync main-process approval timeout with renderer UI cleanup
- When requestApprovalFromRenderer times out, send IPC event
netcatty:ai:mcp:approval-cleared to renderer so stale approval
cards are removed
- Add onMcpApprovalCleared preload bridge for the new IPC channel
- setupMcpApprovalBridge now subscribes to cleared events, removes
timed-out entries from pendingApprovals and notifies clearedListeners
so ChatMessageList drops the stale card
* fix: surface denied inline approvals as errors in UI
- Detect error or denial payloads ("error" string or "ok: false")
returned by tools when the user denies an execution
- Set isError: true on the tool-result message so the ToolCall UI
renders it as a failure (red/rejected) instead of a success (green)
* Add AI support for local terminal sessions
* Fix local AI session metadata and shell safety
* Fix local session cloning and multi-exec errors
* Refactor local shell detection helpers
* Fix local shell helper import path
* Fix CJS imports in renderer
* Use ESM local shell helpers in renderer
* Normalize local shell paths and platform metadata
Security: sanitize command input in resolveCliFromPath, add host allowlist
to streaming endpoint, enforce permission model in MCP server tools, add
safety check to terminal:write IPC, fix broken blocklist regex, remove
renderer-controlled allowedHosts parameter.
Correctness: use sessionsRef for latest state in handleSend, merge
add+update to avoid race condition in streaming, mark assistant message
completed after tool-result, return JSON-RPC error for unhandled ACP
permission requests, add finished guard in ptyExec.
Performance: custom React.memo comparator for ChatMessageList, fix doom
loop detection threshold.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Security:
- Apply checkCommandSafety() in aiExec IPC handler to enforce command blocklist
- Add critical path validation in handleSftpRemove (normalize + blocklist)
- Change rm -rf to rm -r so permission errors surface
Stream lifecycle:
- Abort all active streams on component unmount (abortControllersRef cleanup)
- Wrap stream reader in try/finally with releaseLock() to prevent leaks
- Use refs for inputValue/images in handleSend to stabilize callback identity
State persistence:
- Clear debounce timer before synchronous persist in destructive operations
(clearSessionMessages, deleteSession, deleteSessionsByTarget)
- Add 10MB max buffer guard in ACP client and MCP server NDJSON parsing
i18n:
- Replace hardcoded English strings with t() calls in InlineApprovalCard,
PermissionDialog, ConversationExport, ThinkingBlock, AgentSelector
- Add 23 new i18n keys to en.ts and zh-CN.ts
Misc:
- Remove debug console.log statements in mcpServerBridge
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Fix stale closure: add updateLastMessage to handleApprovalResponse deps
- Use random heredoc delimiter to prevent content corruption when file
contains the literal delimiter string
- Remove dead ensureClaudeConfigDir function from claudeHelpers
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Safety enforcement:
- Command Timeout: use user setting in MCP Server (execViaPty/execViaChannel)
and Catty Agent path (aiBridge execViaPtyForCatty/execViaChannel fallback)
- Max Iterations: read user setting for Catty (stepCountIs) and ACP (via IPC)
- Permission Mode: observer mode hard-blocks write ops in MCP Server dispatch;
Catty SDK tools wired to checkToolPermission(); all synced via IPC on change
Command Blocklist UI:
- Add editable regex pattern list in Settings AI Safety section
- Reset to defaults button, add/remove patterns
- IPC sync to MCP Server on change and on mount
Provider UX:
- ModelSelector rewritten as combobox: type-to-filter with suggestions dropdown
- All providers use unified ModelSelector (modelsEndpoint optional)
- API key passed as auth header for model fetching (Bearer / x-api-key)
- Skip model fetch when no API key (except Ollama)
- Provider toggle is now mutually exclusive (activating one disables others)
- New providers default to disabled (switch off)
- Custom provider supports editable display name
- No-provider error: friendly message shown in chat
i18n:
- Full i18n coverage for Settings AI tab (~55 keys)
- Full i18n for AI chat panel (placeholder, empty state, time, sessions)
- en.ts and zh-CN.ts locale files updated
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Remove chatSessionId from MCP server config env vars so it doesn't
affect the fingerprint calculation. Previously, different chat sessions
in the same workspace produced different fingerprints, causing the ACP
provider to be recreated when switching back to a previous session,
losing all conversation history.
Now the fingerprint only depends on the workspace scope (host session
IDs and MCP port), so providers are correctly reused when returning
to a previous chat session. Each chat session still has its own
provider instance (keyed by chatSessionId in the acpProviders Map).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Fix workspace host discovery: use `activeWorkspace.root` instead of
non-existent `.tree` property, which caused `collectSessionIds` to
always return empty for workspaces
- Isolate AI chat state per-scope (tab/workspace): activeSessionId,
isStreaming, abortController, and inputValue are now keyed by
`${scopeType}:${scopeTargetId}` so different tabs don't interfere
- Add per-chatSession MCP scope metadata to prevent host list mixing
across workspaces, with chatSessionId passed through the full IPC chain
- Store hostname/username in SSH session objects as fallback for MCP
host info when renderer metadata is unavailable
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The MCP server was exposing ALL terminal sessions to AI agents regardless
of which workspace the agent belonged to. Fixed by:
- Track scoped session IDs when updateSessionMetadata is called
- buildMcpServerConfig now auto-uses current scoped IDs when no explicit
scope is provided, setting NETCATTY_MCP_SESSION_IDS env var
- handleGetContext falls back to sessionMetadata keys when no explicit
scopedSessionIds param is passed
This ensures agents only see hosts within their workspace/terminal scope.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>