Open Source CLI · Apache 2.0
You didn't sign up to babysit an AI. Define your agents in TOML, pipe data through them, chain them together. Axe does the work and gets out of the way.
Every time you need something done, you open a chat window, type a prompt, babysit the response, and repeat it again tomorrow.
That's not automation. That's delegation theater — the appearance of AI productivity with none of the actual leverage.
Real automation is silent and repeatable. It runs from cron. It wires into git hooks. It reads stdin and writes stdout. It doesn't need you in the loop.
01 — DEFINE
Drop an agent config in ~/.config/axe/agents/.
Name it, give it a model, point it at a SKILL.md.
That's your agent.
02 — BRING YOUR KEY
Anthropic, OpenAI, Ollama, Gemini — set your API key as an env var.
Axe uses the models.dev format. Zero vendor lock-in.
03 — RUN
axe run <agent> accepts stdin, writes stdout.
Wire it into cron, git hooks, entr, or CI.
Chain agents with pipes.
Agents run in isolated containers, spun up and destroyed automatically. You bring the agent config; we handle the execution environment.
Append-only memory survives between runs. Editable from the dashboard. Agents accumulate context without bloating every prompt.
Start time, end time, status, tokens consumed — every run logged and auditable. Know exactly what ran, when, and at what cost.
TOML and SKILL.md files stored with full version history. Pin a specific version. Roll back when something breaks. Your agents are code.
Role-based permissions across agent configs, skills, and remote execution. Share agents with the team without sharing your API keys.
Per-agent token usage and estimated cost on every run. No surprises on the model API bill. Bundled in all paid tiers.
Run it yourself, embed it, fork it — no license friction. The open source CLI is the full tool for solo developers. getaxe.dev is the hosted layer for teams who want to stop managing infrastructure.