terma) connects your coding agents to Terma, then stamps the commits they produce so agent spend can be traced to the code it shipped.
It is deliberately small. You install it, run terma setup once, and you are done: there is no per-repository step, no file to commit, and nothing to configure per project. Your team’s collection policy, set in Terma, decides which repositories are collected.
Four commands
Every command is safe to run again.
terma setup, so it loads the new configuration.
terma has a few more commands that other programs run for it: your agents and git run terma hook, the relay’s background service runs terma relay, and the hooks deliver their queue with terma spool. They are hidden from terma --help, and you never need to run them yourself.
How it fits together
1
Your team chooses what is collected
An organization admin sets the team’s collection policy in Terma (Team settings → Data collection): every session, or only the repositories it lists, and whether prompts and tool content are included.
2
Setup wires your machine once
terma setup points your agents’ telemetry at a relay terma runs on 127.0.0.1 and writes their machine-wide hooks. It does not touch git’s configuration and writes nothing into a repository’s files.3
Your agents work as usual
A hook claims each session in a repository your team collects. The relay forwards only those sessions, with only the content the policy allows. Everything else never leaves your machine.
4
Commits are stamped
When the policy asks for commit stamping, the first agent session in a collected repository adds two git hooks to its
.git/hooks. Commits of files an agent edited get Agent-Session-Id and Agent-Tool trailers.Global flags
The CLI is open source under the MIT license: github.com/miradorlabs/terma-cli. An agent-facing guide is published at terma.ai/cli/llms.txt.