Files
openclaw/src/node-host/invoke-agent-cli-claude-handler.ts
Peter Steinberger d8d2f83cc1 feat(terminal): open Codex/Claude catalog sessions in a terminal on their owning host
Catalog session rows (sidebar context menu + click), the built-in viewer
header, and a new "Open Codex/Claude sessions in" preference can launch the
native CLI (codex resume / claude --resume) in the operator terminal on the
machine that owns the session.

- Gateway-local sessions spawn through the existing terminal launch policy
  (sandbox/enabled gates preserved) with the resume command in the session cwd.
- Paired-node sessions run through a new seq-ordered node PTY relay: a
  duplex node-host command streams PTY output via node.invoke.progress and
  receives keystrokes/resize via a new node.invoke.input event, behind the
  unchanged terminal.* client protocol (TerminalSessionManager gains a backend
  abstraction; node relay reuses the streaming-invoke controller).
- Owner boundary: each plugin owns its resume command and builds argv from a
  validated thread id; the gateway routes node opens through the node command
  allowlist and plugin invoke policy (no advertisement-only trust), and nodes
  re-verify session eligibility before spawning.
- UI setting catalogOpenTarget + canOpenTerminal capability gate every entry
  point; capability requires the owning host to actually have the CLI.

Node PATH is normalized before command-availability probes, Windows .cmd/.bat
shims spawn via ComSpec, and catalog terminal opens reattach persisted tabs
before opening the new tab.
2026-07-13 22:20:50 -07:00

190 lines
6.3 KiB
TypeScript

import { createExecApprovalPolicySnapshot } from "../infra/exec-approvals.js";
import type { OpenClawPluginNodeHostCommandIo } from "../plugins/types.js";
import type { NodeHostClient } from "./client.js";
import {
decodeClaudeCliNodeRunParams,
type ClaudeCliNodeRunParams,
type ClaudeCliNodeRunResult,
} from "./invoke-agent-cli-claude-params.js";
import { runClaudeCliNodeCommand } from "./invoke-agent-cli-claude.js";
import {
buildSystemRunApprovalPlan,
handleSystemRunInvoke,
resolveEffectiveSystemRunExecPolicy,
} from "./invoke-system-run.js";
import type { NodeInvokeRequestPayload, RunResult, SkillBinsProvider } from "./invoke-types.js";
export type NodeHostInvokeRuntime = {
claudePath?: string;
handleSystemRun?: typeof handleSystemRunInvoke;
signal?: AbortSignal;
pluginCommandIo?: OpenClawPluginNodeHostCommandIo;
};
type ClaudeCliNodeInvokeDeps = Pick<
Parameters<typeof handleSystemRunInvoke>[0],
| "resolveExecSecurity"
| "resolveExecAsk"
| "isCmdExeInvocation"
| "sanitizeEnv"
| "runViaMacAppExecHost"
| "buildExecEventPayload"
> & {
sendErrorResult: (
client: NodeHostClient,
frame: NodeInvokeRequestPayload,
code: string,
message: string,
) => Promise<void>;
sendInvalidRequestResult: (
client: NodeHostClient,
frame: NodeInvokeRequestPayload,
error: unknown,
) => Promise<void>;
sendInvokeResult: (
client: NodeHostClient,
frame: NodeInvokeRequestPayload,
result: {
ok: boolean;
payload?: unknown;
payloadJSON?: string | null;
error?: { code?: string; message?: string } | null;
},
) => Promise<void>;
};
export async function handleClaudeCliNodeInvoke(params: {
frame: NodeInvokeRequestPayload;
client: NodeHostClient;
skillBins: SkillBinsProvider;
runtime: NodeHostInvokeRuntime;
deps: ClaudeCliNodeInvokeDeps;
}): Promise<void> {
if (!params.runtime.claudePath) {
await params.deps.sendErrorResult(
params.client,
params.frame,
"UNAVAILABLE",
"Claude CLI agent runs are unavailable",
);
return;
}
const claudePath = params.runtime.claudePath;
let request: ClaudeCliNodeRunParams;
try {
request = await decodeClaudeCliNodeRunParams(params.frame.paramsJSON);
} catch (error) {
await params.deps.sendInvalidRequestResult(params.client, params.frame, error);
return;
}
const approvalCommand = [claudePath, ...request.argv];
const preparedApproval = buildSystemRunApprovalPlan({
command: approvalCommand,
...(request.cwd ? { cwd: request.cwd } : {}),
...(request.agentId ? { agentId: request.agentId } : {}),
...(request.sessionKey ? { sessionKey: request.sessionKey } : {}),
});
if (!preparedApproval.ok) {
await params.deps.sendErrorResult(
params.client,
params.frame,
"INVALID_REQUEST",
preparedApproval.message,
);
return;
}
const { getRuntimeConfig: getNodeRuntimeConfig } = await import("../config/config.js");
const execPolicy = await resolveEffectiveSystemRunExecPolicy({
cfg: getNodeRuntimeConfig(),
agentId: request.agentId,
defaultSecurity: params.deps.resolveExecSecurity(undefined),
defaultAsk: params.deps.resolveExecAsk(undefined),
requireSocket: false,
});
const approvalPlan = {
...preparedApproval.plan,
policySnapshot: createExecApprovalPolicySnapshot({
file: execPolicy.approvals.file,
agentId: request.agentId,
}),
};
let runResult: RunResult | undefined;
await (params.runtime.handleSystemRun ?? handleSystemRunInvoke)({
client: params.client,
// The command-specific validator is the execution boundary. Approval sees
// every executable argument; prompt/stdin content remains request input.
params: {
command: approvalCommand,
...(request.cwd ? { cwd: request.cwd } : {}),
...(request.env ? { env: request.env } : {}),
...(request.agentId ? { agentId: request.agentId } : {}),
...(request.sessionKey ? { sessionKey: request.sessionKey } : {}),
...(request.systemRunPlan ? { systemRunPlan: request.systemRunPlan } : {}),
...(request.approvalDecision ? { approvalDecision: request.approvalDecision } : {}),
timeoutMs: request.timeoutMs,
},
skillBins: params.skillBins,
execHostEnforced: false,
execHostFallbackAllowed: true,
resolveExecSecurity: params.deps.resolveExecSecurity,
resolveExecAsk: params.deps.resolveExecAsk,
isCmdExeInvocation: params.deps.isCmdExeInvocation,
sanitizeEnv: params.deps.sanitizeEnv,
runCommand: async (approvalArgv, cwd, env, timeoutMs) => {
runResult = await runClaudeCliNodeCommand({
client: params.client,
frame: params.frame,
request,
argv: approvalArgv,
cwd,
env,
timeoutMs,
signal: params.runtime.signal,
});
return runResult;
},
runViaMacAppExecHost: params.deps.runViaMacAppExecHost,
// Agent runs already report through the agent-run stream. Suppress the
// system.run lifecycle side-channel, whose Gateway provenance is scoped
// exclusively to system.run invokes.
sendNodeEvent: async () => {},
buildExecEventPayload: params.deps.buildExecEventPayload,
sendInvokeResult: async (result) => {
if (
!result.ok &&
!request.approvalDecision &&
result.error?.message?.includes("approval required")
) {
await params.deps.sendInvokeResult(params.client, params.frame, {
ok: true,
payloadJSON: JSON.stringify({
approvalRequired: true,
systemRunPlan: approvalPlan,
security: execPolicy.security,
ask: execPolicy.ask,
}),
});
return;
}
if (!result.ok || !runResult) {
await params.deps.sendInvokeResult(params.client, params.frame, result);
return;
}
const payload: ClaudeCliNodeRunResult = {
exitCode: runResult.exitCode ?? 1,
stderrTail: runResult.stderr,
truncated: runResult.truncated,
...(runResult.timedOut
? { timeoutKind: runResult.noOutputTimedOut ? ("idle" as const) : ("hard" as const) }
: {}),
};
await params.deps.sendInvokeResult(params.client, params.frame, {
ok: true,
payloadJSON: JSON.stringify(payload),
});
},
sendExecFinishedEvent: async () => {},
preferMacAppExecHost: false,
});
}