AgentForeman CLI
AgentForeman coordinates coding agents: a chairman clarifies the task, workers execute and the result comes with evidence.
- 54 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- Agents edit code and run processes unattended overnight.
- The system needs authenticated CLI clients and API keys, and token use comes out of your limits.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/agentforemanDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Agents edit code and run processes unattended overnight.
- The system needs authenticated CLI clients and API keys, and token use comes out of your limits.
This is third-party code. Review the repository files before installing.
What it does
AgentForeman is a local orchestration system for coding agents and complex tasks. You connect your CLI clients or API providers, agree on a task with a chairman, and a bounded multi-agent workflow produces a result, critiques it and revises. The base is a versioned task contract: after your confirmation, goals, roles, evidence, checks and execution limits are frozen. Modes are cooperative work with an integrator and independent review, adversarial improvement through criticism, revision and retest, multi-model councils with anonymous peer review, and overnight experiments where Codex edits an allowed scope and an independent evaluator checks the checkpoint. State lives in SQLite, and token accounting is kept per connection.
Who it is for. Researchers and developers who run long experiments and want to split work between several agents with checked results.
Good fit when
- An overnight experiment must run: Codex changes code, a process trains a model and an independent evaluator checks the result.
- A task is first agreed with a chairman and only then executed.
- Data analysis needs a council of several models with anonymous peer review.
- You want token use per connection and task over 1, 7 and 30 days.
Not a fit when
- You do not have three tested connections for adversarial tasks, as cooperative ones need four.
- You need a ready cloud service, since the system runs locally.
- The task is small and a single agent suffices without orchestration.
Example request
Agree a task with me: cut test time by thirty percent. Propose goals, agent roles, checks and limits and begin execution only after my confirmation.Limitations
It needs Python 3.11 or newer, Git and at least one authenticated CLI client or API connection. Automatic adversarial tasks require at least three tested connections and cooperative ones four, while several models on one account share its limit. The interface currently uses Chinese labels while the README is in English, and screenshots show demonstration providers. Unknown token use stays unknown and is never counted as zero. Allocation percentages are targets and hard budgets do not exist. Autonomous overnight code edits must be limited to an allowed scope.
How to disable. Stop the worker processes and delete the project folder created by `init` along with the repository clone. Disconnect CLI and API connections in the panel or remove keys from the environment.
FAQ
How many connections are needed?
At least three tested identities for automatic adversarial tasks and at least four for cooperative ones. Models on one account share its limit.
What is the chairman?
An agent that helps define a versioned task contract. After your confirmation, goals, roles, checks and limits are frozen.
How does it start on macOS and Linux?
Create a virtual environment, install with `pip install -e ".[demo,desktop]"`, run `python -m agentforeman init ./my-experiment` and then `ui`.
What happens when a quota runs out?
Recognized quota resets become persistent waiting periods with no model calls, and state lives in SQLite and survives restarts.
Which pages does the interface have?
Tasks with the chairman conversation, subscriptions and models, and token usage. Advanced tools expose experiments, councils and diagnostics.
Discussion
No questions or reviews yet.
Related
CLI toolsForeman watches coding agents and decides whether to continue, steer, stop, retry or verify their work, so a person need not.
A bridge between local CLI agents: Codex, Cursor or Claude Code delegate work to Antigravity, Grok, Kimi and seven more.
A multi-agent harness that verifies results by artifacts: stop hooks, an independent judge and an event log for Claude Code, Codex, Cursor and more
A layer over coding agents that assembles a temporary specialist team per task inside Claude Code, Codex, Cursor and Gemini
A Go coding agent for open models: it turns a chat into tasks on separate branches, checks and merges the result and runs all projects in one window
Shared infrastructure for Claude Code, Codex, Antigravity CLI and OpenCode: a task lifecycle, skills and sandboxes.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| AgentForeman (this card) | no | 54 | ||
| CLI Agent Orchestrator | no | 1,548 | difference: `cli-agent-orchestrator-cao` runs agents in sessions, while AgentForeman adds a task contract, verification and overnight experiments | |
| Ralph Orchestrator | no | not required | 3,203 | difference: `ralph-orchestrator` loops one agent, while AgentForeman splits work across several connections with independent review |
README in short
The README describes a three-page interface, work modes, a quick start for Windows, macOS and Linux, requirements and documentation. Screenshots use demonstration providers.