The stupid coordination tool
Tasks and architecture decisions in your repo, behind one CLI any coding agent can call.
An agent can read every line of your code, but not your tracker, your wiki, or
the thread where you decided that sessions must never be self-contained JWTs.
Ank keeps those decisions and the work in .ank/, attached to the code they
constrain, and serves them through one CLI.

npm install -g @haksolot/ank # the binary for your platform
npx skills add haksolot/ank # the skills, for your agentLinux, macOS and Windows; needs git 2.34 or newer. The shell installers and
every other route are in Install. The version is 0.x on purpose: the loop and
the exit codes are specified, the storage format is not yet.
ank context <path> # what binds this perimeter, and what is takeable
ank claim <id> # take a task and freeze its criterion
ank show <id> # the entity whole
ank log "<message>" # what you learned, while you work
ank done # run the declared verifiers and record the proof
ank release --reason "<why>" # hand the task backThe quickstart walks it from ank init onward, with real output.
- Not a tracker. No cycles, estimates, velocity or roadmap.
- Not a wiki. Only what is actionable or binding for an agent goes in.
- Not a security boundary. It protects against drift, not against an attacker.
The documentation is published at https://haksolot.github.io/ank/, built from
docs/ on every merge.
- Using ank: Install, the quickstart, proof and
verifiers, claims and
identity, reading
ank check, CI, multi-agent work - Reference: exit codes, the file format, environment variables, the specification
- Integrating: the machine surface, the MCP server, the watcher
- Maintaining: ratifying, releasing, CI jobs
- Contributing · Security · Code of Conduct
Apache-2.0. See LICENSE.