CLI

`trailboss` puts your workspace in your terminal: issues, live releases, agent logs, knowledge, and agents running on your own machine.

Sign in

Create a token in Settings → API tokens, then:

trailboss login

Paste the token when asked (it isn't echoed), or pipe it in with --token-stdin. trailboss whoami shows who it belongs to; trailboss logout forgets it. In CI, set TRAILBOSS_URL and TRAILBOSS_TOKEN instead.

Issues

trailboss issues --status in_progress
trailboss issues create "Retry webhooks with backoff" -p high --label payments
trailboss issue TB-142

Follow work as it happens

trailboss watch release rel_…    # stages, gates and approvals, live
trailboss watch project prj_…    # a project on autopilot
trailboss logs prj_… -f          # the agent log, followed

watch release exits 0 when the release is live and 1 if it doesn't get there, so it fits in a script.

Knowledge

trailboss knowledge sources
trailboss knowledge add-docs ks_… ./docs
trailboss knowledge search "how do we roll back?" --agent

--agent shows exactly what an agent would be given for the question.

Agents on your machine

Run jobs on your computer against a local model, any OpenAI-compatible endpoint (Ollama by default):

trailboss local check --model qwen2.5-coder:7b
trailboss local --model-url http://localhost:11434/v1

local check runs one job end to end so you know it works. The first time, the CLI shows the workspace's job-signing key and remembers it; jobs signed with any other key are refused. If the model needs a key, name the environment variable that holds it with --model-key-env.

For scripts

Every command takes --json. Exit codes: 0 ok, 1 failed, 2 the command was used wrong.