Candidate first
Learning, Skill Creator, Improver, and Sync cannot overwrite active Skills.
Verified CLI · v0.166.65
Your terminal,
your AI workbench.
175 top-level commands covering the full lifecycle from knowledge base to on-chain governance. Zero GUI dependency, plain text I/O — pipeable, scriptable, CI-ready.
npm latest is Agent Platform 0.166.65, tagged v-npm-0-166-65 at 17509017a2. Ordinary Agent chat works without an evolution deployment; configured governance still validates authenticated ingress and never downgrades on failure. Automatic promotion remains HOLD. TypeScript Agent SDK: 0.2.11.
install
$ npm i -g chainlesschain@0.166.65
$ yarn global add chainlesschain@0.166.65
$ pnpm add -g chainlesschain@0.166.65
$ bun add -g chainlesschain@0.166.65
# Aliases: chainlesschain / cc / clc / clchain
$ cc --help
CLI 0.166.65
Keep long tasks moving.
Interactive streamed tasks can continue beyond the old 50-model-call default. Large-file reads keep their exact position through compaction, reuse unchanged pages and refresh after edits. Slow commands keep the IDE session responsive.
Public CLI 0.166.65 allows an explicit bounded read of a known target during recovery and separates synthetic pauses from real file, command, and remote observations, so recovery guidance cannot trigger its own repeated-read or identical-output guard. Genuine repetition, policy rejection, global no-progress, and six-real-attempt limits remain fail closed. The exact-SHA CLI CI/Strict three-OS gates, OIDC provenance, public registry, and fresh-install readback all passed.
Read the settings and upgrade guide → Current in 0.166.65
Governed Skill and knowledge evolution
Workbench unifies candidates, evidence, version comparison, human review, and rollback projections. Skill Retrieval ranks by digest, index witness, and real outcome evidence. Encrypted knowledge conflicts publish only through an authenticated merge plan.
Evolution Workbench
Review and rollback effects persist across restarts. Completed effects are reconciled; unexecuted plans remain deferred.
Evidence-ranked retrieval
cc skill search supports four sources, hybrid retrieval, canonical digests, and verified outcomes.
Encrypted knowledge merge
Revocation follows multi-hop cross-Wiki provenance, with distinct candidate quarantine, rejection, Skill rollback and Wiki tombstone effects.
Target-matrix eval
Target, grader, safety, and verifier roles are separated; every required cell must pass.
Recoverable release
Marketplace installation materializes a verified candidate; rollout and revocation require current state and independent receipts.
Signed deployment config
cc evolution deployment verifies and atomically saves an owner-only profile. Environment overrides win; auto-promotion remains HOLD.
cc evolution deployment init-test/replace-test/status/configure/enable/disable, Workbench, knowledge, and Skill search are public. Before production, the CLI can generate clearly marked local TEST signing credentials and later rotate safely to the managed descriptor/trust root. This grants no review or release authority; production KMS/PKI/witness/grader remain operator-provided. Read command examples and limits →
All 175 top-level commands
31 categories 01 System Management 8 items ▾
cc setup Interactive setup wizard cc start Launch desktop app cc stop Stop the app cc status Show status cc services up Start Docker dependencies cc config list View current config cc update Check for updates cc doctor Diagnose runtime environment 02 Headless Commands (no GUI) 36 items ▾
cc db init Initialize database cc note add Create note (with tags) cc note list List notes cc note search Full-text search cc chat Interactive AI chat (streaming) cc chat --agent Agentic mode (read/write files) cc ask One-shot AI query cc llm models List installed models cc llm test Test LLM connectivity cc agent Claude Code-style agent session cc agent -p Headless one-shot run (claude -p parity) cc exec -p Governed one-shot Agent Kernel facade cc agent --think / --ultrathink Anthropic extended thinking (+ --thinking-budget) cc agent --image Multimodal image input (vision models) cc agent web_search Built-in pluggable search tool (keyless backends incl.) cc agent -p --json-schema Structured output: reply validated against a JSON Schema with auto-retry cc agent REPL steering Queued input / Esc interrupt / /rewind double-Esc / ! bash passthrough / # memorize / TAB completion cc init Inventory the folder into a cc.md memory file (claude /init parity; --ai agent refine) cc memory files Show the project-memory chain the agent auto-loads (cc.md/CLAUDE.md/AGENTS.md + path-scoped rules) cc session export / search Export transcripts as Markdown / full-text search across sessions cc cost Token spend accounting cc checkpoint File-state snapshot / rewind (git-plumbing dual engine) cc compact Session context compression (/compact parity) cc loop Recurring / self-paced prompt & command runner cc goal Cross-session persistent goals / OKR cc permissions .claude/settings.json allow/ask/deny rules cc auto-mode eval [--json] Run the dangerous-operation classifier corpus and CI gate offline, without command execution or file/network I/O cc output-style list Output style personas (explanatory/learning) cc statusline preview Custom / built-in context-usage status line cc mcp login Remote MCP server OAuth (PKCE) cc ide list IDE bridge: discover & connect editor MCP server cc command list / run Custom slash-command macros (.claude/commands/*.md, $ARGUMENTS/!bash/@file expansion) cc skill list List all skills (4 layers) cc skill run Run a specific skill cc skill sync-cli Generate / update 9 CLI Skill Packs cc skill add Create a custom project skill 03 AI Intelligence Layer 4 items ▾
cc search BM25 hybrid search cc tokens show Token usage tracking cc memory add Persistent memory cc session list Session management 04 Knowledge & Content 6 items ▾
cc import markdown Import a Markdown directory cc import evernote Import Evernote backup cc import pdf Import PDF cc export site Export as static HTML cc git status Git integration status cc git auto-commit Auto-commit changes 05 MCP & External Integration 11 items ▾
cc mcp servers List MCP servers cc mcp add Add an MCP server cc mcp serve Expose local file tools as an MCP server (root-confined + Bearer auth) cc mcp tools List available tools cc ide status / doctor IDE bridge discovery + diagnostics (VS Code on Open VSX, JetBrains plugin approved + listed) cc agent --ide Auto-connect inside the editor: selection shared per prompt, diagnostics fed back, edit approval as native diffs (hunk-level partial accept), Jupyter kernel execution, guided LLM setup (both editors), sidebar chat panel (streaming replies, plan approval, Approve/Deny cards for risky actions, Esc interrupt, session resume + /sessions picker) cc browse fetch Fetch a web page cc browse scrape Scrape elements cc llm providers List 10 LLM providers cc llm switch Switch active provider cc instinct show View learned preferences 06 Security & Identity 9 items ▾
cc did create Create DID identity (Ed25519) cc did list List all identities cc did sign Sign with default DID cc encrypt file AES-256-GCM file encryption cc decrypt file Decrypt file cc auth roles List RBAC roles cc auth check Check permissions cc audit log Recent audit events cc audit stats Audit statistics 07 P2P / Blockchain / Enterprise 11 items ▾
cc p2p peers List P2P peers cc p2p send Send encrypted message cc p2p pair Pair devices cc sync status Sync status cc sync push / pull Push & pull data cc wallet create Create wallet cc wallet assets List digital assets cc wallet transfer Transfer assets cc org create Create organization cc org invite Invite members cc plugin list / install Plugin management 08 Init & Persona 5 items ▾
cc init Interactive project init cc init --template code-project Code project template cc init --template medical-triage Medical triage template cc init --template ai-doc-creator AI doc creator template cc persona show / set / reset Persona management 09 Cowork Multi-Agent 8 items ▾
cc cowork debate Multi-perspective code review cc cowork compare A/B option comparison cc cowork analyze Code analysis cc cowork status Cowork status cc team run --tasks <file> Declared dependency DAG with budgets, leases, worktrees, and recoverable state cc team queue init / worker / finalize Shared-filesystem distributed queue v1 with CLI-owned authority digests, leases, and CAS fences cc team graph inspect / diff / eval Read-only GraphRun time travel, root cause, diff, and CI thresholds cc team_send / receive / ack / followup 0.166.3 real-Team-child messaging bound to attempt, lease, and fence authority 10 AI Core: Hooks / Workflow / Memory / A2A 4 items ▾
cc hook list / add / run Hook management cc workflow create / run / templates Workflows cc hmemory store / recall / consolidate Hierarchical memory cc a2a register / discover / submit Agent-to-agent protocol 11 Security & Evolution 6 items ▾
cc sandbox create / exec / audit Sandbox cc evolution workbench / knowledge Evolution review and knowledge merge cc skill search Evidence-ranked Skill retrieval cc evomap federation EvoMap federation cc evomap gov Governance vote cc dao propose / vote / delegate DAO governance 12 Blockchain & Compliance 8 items ▾
cc economy pay / market / nft Economy layer cc zkp compile / prove / verify Zero-knowledge proofs cc bi query / dashboard BI engine cc compliance evidence / report Compliance evidence cc dlp scan / incidents Data loss prevention cc siem targets / export SIEM integration cc pqc keys / generate / migrate Post-quantum cryptography cc nostr / matrix / scim / terraform Protocol bridges 13 Server & Web Panel 5 items ▾
cc serve WebSocket server (18800) cc serve --app-server stdio JSON-RPC product integration: Thread / Turn / Item / Approval cc serve --app-server --app-server-websocket Experimental App Server WS: fixed path/subprotocol, mandatory token, remote TLS cc ui Web admin panel (18810) cc stream NDJSON streaming output 14 AI Orchestration 3 items ▾
cc orchestrate "task" Auto-detect agent backend cc orchestrate detect Detect Claude Code / Codex cc orchestrate --webhook Webhook server (WeCom / DingTalk / Lark) 15 Communication Bridges & Social 5 items ▾
cc nostr keygen / publish / dm Nostr NIP-01/04/09/25 cc matrix login / rooms / send Matrix threads & spaces cc activitypub actor / publish / follow ActivityPub C2S federation cc social graph add-edge / neighbors Social graph (directed weighted) cc social analyze / detect-lang Multi-language topic classification 16 Stress Testing & Ops 5 items ▾
cc stress run / analyze / plan Stress test + capacity planning cc sla create / check / report SLA management (gold/silver/bronze) cc reputation observe / score / anomalies Reputation scoring + anomaly detection cc tech analyze / detect / recommend Tech-stack analysis + anti-pattern detection cc hardening config-check / deploy-check Config audit + deploy preflight 17 Autonomous Dev & Governance 5 items ▾
cc dev start / phase / review / adr Autonomous developer (L0–L4 autonomy) cc collab propose / vote / execute Collaboration governance + task assignment cc governance create / vote / tally Community governance (proposals + voting) cc learning reflect / synthesize Autonomous learning system cc perm / fflag / rcache / seshu … V2 governance command family (27 developer surfaces: governance/lifecycle state machines, see docs) 18 Marketplace & Economy 3 items ▾
cc marketplace publish / list / stats Skill marketplace (publish + invocation stats) cc incentive mint / transfer / leaderboard Token incentive (7 contribution types) cc crosschain bridge / swap / send Cross-chain interop 19 Knowledge Graph & Recommendation 3 items ▾
cc kg add / query / reason Enterprise knowledge graph (multi-hop reasoning) cc recommend profile / recommend / feedback Smart recommendation system cc tenant create / subscribe / check-quota Multi-tenant SaaS engine 20 Video Editing Agent 3 items ▾
cc video edit End-to-end video editing (CutClaw) cc video deconstruct / plan / assemble / render Step-by-step workflow cc video assets list / prune Asset cache management 21 Privacy Computing 3 items ▾
cc privacy create-model / train Federated learning (FL) cc privacy create-computation / submit-share Secure multi-party computation (MPC) cc privacy dp-publish / he-query Differential privacy / homomorphic encryption 22 Code Generation & DB Evolution 4 items ▾
cc codegen generate / review / scaffold Code Generation Agent 2.0 cc codegen stats Generation statistics dashboard cc dbevo register / up / down Database migration management cc dbevo analyze / apply Slow-query analysis + index suggestions 23 AIOps 4 items ▾
cc ops baseline-update / detect Metric baselines + anomaly detection cc ops incident-create / ack / resolve Incident lifecycle management cc ops playbook-create / toggle / record Playbook auto-remediation cc ops postmortem Postmortem report 24 Multimodal Perception & Federation 5 items ▾
cc perception record / results / query Perception results & cross-modal search cc perception voice-start / voice-status Voice session state machine cc perception index-add / context Multimodal indexing & context cc federation register / breakers / check Circuit breakers + health checks cc federation pool-init / pool-stats Connection pool management 25 NL Programming & Quantization 5 items ▾
cc nlprog classify / extract / detect-stack Intent / entity / stack detection cc nlprog translate / refine NL → structured spec translation cc nlprog convention-add / conventions Project coding conventions cc quantize create / start / complete GGUF/GPTQ quantization tasks cc quantize levels / gptq-bits Quantization level catalog 26 Runtime & IPFS & Multimodal 5 items ▾
cc runtime plugin-load / hot-update / rollback Plugin hot update + rollback cc runtime state-set / state-get / state-list LWW state sync cc runtime profile / health / metrics Runtime monitoring cc ipfs add / get / pin / gc IPFS content I/O + GC cc mm session-create / fuse / generate Multimodal fusion & generation 27 Agent Network & Identity 6 items ▾
cc anet did-create / register / discover Agent node registration + discovery cc anet auth-start / credential-issue Authentication + W3C VC credentials cc anet task-route / rep-update Reputation-weighted task routing cc didv2 create / cred-issue / present DID 2.0 + VC/VP cc didv2 recover-start / roam / rep-aggregate Social recovery + roaming + reputation cc sso create / login-url / link SSO SAML/OAuth2/OIDC + DID bridging 28 Pipeline & Automation & Plugins 10 items ▾
cc pipe create / start / approve / deploy 7-stage development pipeline cc pipe rollback / monitor-record Deployment rollback + monitoring cc auto create / activate / execute Workflow automation (12 SaaS connectors) cc auto add-trigger / fire-trigger Webhook/Schedule triggers cc routine run / trigger / runs Unified Scheduler: routines, manual triggers, and run history cc agenda run --watch Resident Agenda wakeup/cron scheduling cc cowork cron run Durable Cowork Cron scheduling and recovery cc automation run-scheduled / center-projection Unified Automation Scheduler and governed IDE control plane cc eco register / review / publish Plugin ecosystem (AI review + sandbox) cc perf collect / evaluate / apply Performance tuning suggestions 29 Personal Data Hub & Mobile Companion 9 items ▾
cc hub list-adapters 92 collection contracts plus current readiness/setup reason cc hub sync-adapter <name> One source via direct sync, file, directory, transient Cookie/OAuth, or ADB cc hub sync-all Run only currently ready sources and summarize success/empty/skipped/failure cc hub ask Ask against local data cc pair preflight LAN pairing preflight diagnostics cc pair token Issue pairing token cc notification send Push notifications to paired mobile devices cc android Android device bridge (15 subcommands: system-data / bridge-direct / PDH collection) cc project list / create Desktop projects (shared SQLite, mobile-sync via Phase 3d) 30 MTC Federation & Multisig & Packaging 8 items ▾
cc mtc federation create Merkle Tree Certs M-of-N federation cc mtc federation sync libp2p / filesystem gossipsub auto-discovery cc multisig propose / sign / finalize M-of-N multisig proposals (beyond MTC publisher_signature) cc multisig sweep Expire-proposal cleanup cc pack Bundle current directory into a standalone executable cc pack --project Bake project skills / persona / whitelist into the binary cc pack check-update Probe OTA manifest cc cli-anything scan / register Discover & register Agent-native CLI tools as Skills 31 Decentralized Network & Low-Code 5 items ▾
cc lowcode create / publish Low-code application platform (15+ components) cc inference register / route Decentralized inference network (Phase 67) cc trust verify / score Trust & security (Phase 68-71) cc fusion ... Protocol fusion & AI social (Phase 72-73) cc infra ... Decentralized infrastructure (Phase 74-75) Agent mode · cc agent
One agent in your terminal,
plus a whole suite of commands around it.
cc agent (interactive REPL) and
cc agent -p (headless single-turn, for CI / pipelines) match Claude Code's claude / claude -p:
read/write files, run commands, search the codebase, call skills, search the web — with a permission gateway, checkpoint rollback, a memory system and bridge auto-discovery built around it.
cc agent -p Headless single turn + text/json/stream-json output + --max-turns / --permission-mode
cc exec -p Governed one-shot facade reusing Agent Kernel permissions, sandboxing, budgets, and events
cc permissions .claude/settings.json allow/ask/deny rules + credential-read guard
cc checkpoint Auto-snapshot inside a git repo; one-click restore when an edit goes wrong
cc cost Estimated token $ spend, split by provider/model; /cost in the REPL
cc loop Repeat a prompt or command on a fixed / adaptive interval with stop conditions
cc goal Persistent cross-session goals / OKRs the agent advances each turn
cc mcp / ide / pdh Feed MCP servers, your editor and an Android device into the agent loop; large tool surfaces auto-enable Tool Search with deferred schemas, saving 10-30% context
cc team run Run a declared task DAG; one local runner is validated at 10,000 tasks and up to 64 workers with durable budgets and recovery state
cc team queue Coordinate multi-process workers on a supported shared filesystem through queue v1; authority digests, leases, and CAS reject stale writes
cc team graph inspect / diff / eval: read-only GraphRun projections, time travel, blocked roots, and CI gates, with distinct Task, Agent, and Trace graphs
cc serve --app-server Stable stdio product entry plus an experimental, token-protected, remote-TLS, fully bounded WebSocket
cc auto-mode eval Audit dangerous-operation classification against an immutable 145-case offline corpus; a failed gate exits nonzero and never executes evaluated commands
cc output-style Named personas layered onto the system prompt (explanatory / learning)
--json-schema Structured output: the reply is validated against a JSON Schema, retried on failure
CLI Skill Packs · 9 domain packs
One command,
all skill bindings generated.
cc skill sync-cli generates 9 domain Skill Packs with 101 built-in tests — ready to use out of the box.
cli-knowledge-pack direct note / search / session — knowledge commands
cli-identity-pack direct did / auth / encrypt — identity & encryption
cli-infra-pack direct db / services / doctor — infrastructure
cli-ai-query-pack llm-query chat / ask / llm — smart Q&A
cli-agent-mode-pack agent agent / cowork — multi-agent
cli-web3-pack direct wallet / nft / dao — on-chain ops
cli-security-pack direct audit / sandbox / dlp — security & compliance
cli-enterprise-pack direct org / scim / terraform — enterprise
cli-integration-pack hybrid mcp / browse / nostr — integration bridges
Custom slash commands · cc command
Write your go-to prompts as files,
summon them with a slash.
Every Markdown file in .claude/commands/*.md (project)
or ~/.claude/commands/ (personal) becomes a
/command — file-format compatible with Claude Code.
Supports $ARGUMENTS / $1-$9 parameters,
inline !`bash` execution and @file expansion;
prompts from connected MCP servers auto-register as /mcp__server__prompt too.
Run headless via cc command run <name>. Unlike AI-triggered Skills,
these are deterministic prompt macros invoked explicitly by you.
Reliability · client/transport hardening
Flaky networks, dead servers, truncated downloads —
it errors when it should, times out when it should.
cc's network / IO client layer is continuously hardened against Claude Code CLI: write_file / edit_file verify the on-disk byte count (a silent truncation on a network / cloud-synced drive no longer reads as success); the MCP client surfaces tools/list failures, fails fast on a dead stdio server (no 30s hang) and times out HTTP requests; cc mcp serve bounds request body + timeout + errors; cc setup verifies Content-Length + a stall timeout. Every guard ships with tests.
Runtime boundaries are covered too: the credential agent forwards only explicit session identifiers while filtering unknown credential-bearing environment variables; async hooks reap the whole process tree on timeout or session shutdown, with a descendant-snapshot fallback when Windows taskkill is denied; and CHAINLESSCHAIN_HOME isolates configuration, sessions, and state in one exact directory.
CLI 0.166.65 extends strong process boundaries to plugin hooks, MCP, LSP, monitors, PTYs, background tasks, and native bins. Linux pins the plugin tree by descriptor before entering a fixed bubblewrap policy; Windows AppContainer binds the target handle and policy digest. Copied, replayed, or stale contracts, dynamic loaders, and missing isolation primitives fail before launch.
P2-14 managed rollback covers only declared workspace write paths owned by the Process Broker. It establishes durable checkpoints at accepted boundaries, applies fenced rollback after failure, cancellation, or timeout, and reports full, partial, or none coverage honestly. Unmanaged writers, additional roots, and external-system side effects remain outside the rollback guarantee.
P2-16 Agent Teams uses local state v6 as authority and adds a separate distributed queue v1 for supported shared filesystems. Queue authority digests, leases, CAS, and wall-clock fences prevent stale workers or late results from publishing. A 10,000-task DAG with up to 64 local workers is the validated boundary for one local runner; the shared-filesystem queue is not consensus and is not network-partition tolerant.
CLI 0.166.65 retains real cc team --agent child messaging, TeamMailbox v3, custody handoff, and the state-backed SessionMessageFabric. All 37 Schema-generated events now have payload discriminated unions and cross-client causal conformance, while transports continue preserving unknown future events losslessly.
The current CLI carries PDH 0.4.60. Source aliases, raw observations, field-level conflict decisions, rewritten references, and final entities commit in one transaction; analysis skills, entity resolution, and embedding selection preserve host model-governance failures. Desktop, Web Panel, and CLI share one live-collection protocol; transient Cookie/OAuth inputs are not persisted. Native SQLite remains optional, and partial reads or incomplete pagination retain the previous checkpoint for a safe resume.
Agent Platform 0.166.65 publishes canonical Agent Protocol, CC App Server, Agent Kernel, durable Graph Kernel, Graph trace/diff/eval, the governed cc exec facade, real-Team-child messaging, and Agent SDK 0.2.11. Exact SHA 17509017a2 passed three-OS CLI CI, Strict Sandbox, Trusted Publishing, and independent registry/provenance readback; the preceding `0.166.5` 1,800-second App Server overload/RSS soak and npm/PyPI SDK install readback remain the corresponding focused evidence.
A GraphRun is the envelope for run identity, authority, budget, and revision: the Task Graph owns dependencies and the ready frontier, the Agent Tree describes dynamic execution topology, and Artifact/Trace graphs are read-only evidence projections. Occurrence admission success or an empty ready set does not imply Graph success. The CLI authoritative entry cut over in 0.166.65@17509017a2; fixed-capability Desktop/VS Code App Server surfaces remain separate migration planes. PM recovery source fbc828c622 proves only read-only readiness and test evidence; IDE release source 063d491dab aligns with 0.166.65 without authorizing Desktop native distribution or automatic promotion.
Updates and release status · npm chainlesschain
Every release is on the record —
read it with cc changelog.
The CLI ships on its own npm alpha track (currently 0.166.x). The list distinguishes source candidates, npm latest, and fully gated builds; registry upload alone is not a completed release gate.
once installed, cc changelog shows the full per-version notes offline
(--all, <version>, --json).
- v0.166.65 npm latest · public readback
- v0.166.64 historical
- v0.166.63 historical
- v0.166.62 historical
- v0.166.61 historical
- v0.166.60 historical