← All docs

CLI Reference: mergen-server

One binary operates the local runtime, the policy gate, and every recovery path. Commands below are grouped the way you'll actually reach for them: get running, operate day-to-day, manage policy, configure, and, once you're past day one, report and audit.

The primary binary is mergen; mergen-server remains a compatibility alias. Run mergen help, mergen help --all, or mergen help <command>.


Get started

Command What it does
mergen-server Zero-config demo: blocks a real destructive command, no account needed
mergen-server setup Installs Mergen, configures your detected AI IDE, and starts the local runtime. No questions asked.
mergen setup --terminal Terminal-only setup. Does not read or modify IDE configuration; use mergen run -- <command> to gate commands explicitly.
mergen-server setup --full The same, plus guided Slack, PagerDuty, GitHub, and git-hook connection
mergen-server setup --ide cursor Configure one specific IDE (cursor | vscode | claude-code | windsurf)
mergen-server start Starts the local runtime (production mode)
mergen-server stop Stops this workspace's running server; also the sanctioned recovery command if the gate is ever wedged
mergen-server service install Supervises the server with your OS (launchd/systemd): starts at login, restarts on crash
mergen-server login Authenticates the CLI: only needed for cloud/team features
mergen-server dashboard Opens My Machine in your browser: this machine's gate activity across every connected IDE
mergen-server governance Opens My Organization on mergen.app: org-wide policies, members, and billing (human-only; no agent can reach this)

Operate & recover

These work even when the gate itself is unreachable: they're recognized directly by the Claude Code hook without a network round-trip.

Command What it does
mergen-server status Live snapshot: server health, buffer size, errors, gate activity
mergen doctor Core health check: runtime, policy, local protection, and orphaned processes
mergen doctor --integrations Also validate optional external integrations
mergen-server doctor --fix Repairs what it finds: prunes stale records, kills confirmed orphans, repairs the hook, starts a clean instance
mergen-server doctor --security Hardening posture checklist: what an agent with filesystem access could forge on this install, and how to close it
mergen-server approve <token> Approves a held tool call by token (honored once, for 10 minutes)
mergen approvals Lists and resolves pending approvals owned by the current workspace
mergen-server approve --last Approves the most recent pending hold, no token to copy
mergen-server trust-reset Resets this session's scrutiny-tier reputation to normal, if every command is suddenly being held
mergen-server panic on --minutes 15 --reason "..." Operator escape hatch: passes everything unenforced for a bounded window (max 4h), fully audited
mergen-server exec -- <cmd> Runs a command through the full gate, then executes it: the explicit entrypoint for anything outside an IDE
mergen run -- <cmd> Preferred terminal workflow; alias of exec
mergen check -- <cmd> Evaluate a terminal command without executing it
mergen events [--follow] Unified recent/live enforcement stream
mergen-server gate-check -- <cmd> Evaluates only, no execution; this is what the Claude Code PreToolUse hook calls

Policy

Command What it does
mergen-server policies Lists active rules with action and trigger counts
mergen-server promote Promotes the observation-window findings into active enforcement
mergen-server policy-push Saves the live server's policy to .mergen/policy.json, for git
mergen-server policy-pull Applies .mergen/policy.json to the live server
mergen-server policy-diff Shows what would change between the file and the live server
mergen-server policy-simulate <file> Replays a candidate policy against historical traffic: see what would newly block or hold
mergen-server policy-test [file.json] Runs {input, expectedVerdict} test cases, no server required, CI-usable
mergen-server policy-versions Lists restorable full-policy snapshots
mergen-server policy-rollback <id> Restores a prior snapshot (immutable rules always survive)
mergen-server scan <url> Scans an MCP server for injection vectors and risk signals before you connect it
mergen-server test-safety Runs the adversarial bypass test suite against your live gate

See Policies for the rule format and condition fields.


Configure

Command What it does
mergen-server config Inspects every effective setting: value, source, and any ACTIVE settings that weaken enforcement
mergen-server config set K V Writes an org-policy setting to the signed config file (guarded keys need --acknowledge-risk)
mergen-server config modules Shows the core/module/integration manifest with live enabled state
mergen-server config profile team Materializes a whole policy profile at once (solo | team | enterprise | fintech)
mergen-server connect <name> Guided integration setup: slack | pagerduty | github | datadog | jira | linear | email | sync-plane
mergen-server agent-register <id> Issues a signed identity token for an agent profile
mergen-server agent-list Lists issued agent identity tokens on this machine
mergen-server test Validates the installation end-to-end
mergen-server test hitl End-to-end HITL check: synthetic hold → Slack delivery → approve round trip
mergen-server ci CI smoke test (exit 0 = healthy)
mergen-server --version Shows version
mergen completion <bash|zsh|fish> Prints a shell-completion script for the consolidated command tree

Full environment-variable reference (151 settings, generated, always current): Configuration Reference.


Reporting & insights

Command What it does
mergen-server impact-report Prints the board-deck metrics: autonomous resolution rate, MTTR, blocks prevented
mergen-server export-risk-report Prints a CISO-facing security risk report
mergen-server verify-log Verifies the Agent Blunder Log's hash chain (tamper-evident audit)
mergen-server trace-replay <traceId> Reconstructs the full gate-event + blunder-log timeline for one cross-process trace

Run mergen-server help for the complete list, including advanced commands (watch, explain, guard, backfill, pr) used less often day-to-day.

Last updated September 9, 2026