meridian-cli 0.2.4

Orchestrate CLI + MCP server for multi-model agent workflows

meridian

Orchestrate AI coding agents across Claude Code, Codex, and OpenCode. Each model runs through its native harness — meridian coordinates between them.

Early development. Meridian is not stable. Expect breaking changes in any release. If you need a stable workflow, this project is not ready for you yet.

v0.1.0 Migration

v0.1.0 removes all legacy state format support in the state layer.

If you are upgrading from any pre-0.1.0 build, run:

rm -rf ~/.meridian

This is a clean break. There is no automatic migration.

Why

Claude Code is (supposed to be) the best way to run Claude. Codex CLI is often the best way to run GPT. Each provider's harness is optimized for their models — sandboxing, tool use, context handling, everything (supposedly).

But you can only use one at a time.

Meridian lets agents spawn other agents across harnesses. The right model for each task, through the right runtime, with clean context per spawn.

Install

uv tool install meridian-cli

Other methods

pipx install meridian-cli
pip install meridian-cli

From source:

git clone https://github.com/haowjy/meridian-cli.git
cd meridian-cli
uv tool install --force . --no-cache --reinstall

You need at least one harness installed: Claude Code, Codex CLI, or OpenCode.

Set Up a Project

meridian init by itself only bootstraps Meridian project config:

meridian init

To initialize Mars package content in the same step, use setup flags:

meridian init --add haowjy/meridian-dev-workflow --link .claude

For first-run onboarding, bootstrap can do setup + launch in one command:

meridian bootstrap --add haowjy/meridian-dev-workflow --link .claude

--add / --link setup flags cannot be combined with --dry-run.

Usage

Launch an interactive session:

meridian

Or spawn agents directly:

# Code on Codex, review on Claude
meridian spawn -a coder -p "Add rate limiting to the API endpoints"
meridian spawn -a reviewer --from p1 -p "Review the rate limiting implementation"

# Check on work
meridian spawn list
meridian spawn show p1

Agents route to their configured model and harness automatically. Each spawn gets a fresh context window with only the context it needs.

Architecture

graph TB
    User([You]) --> Primary["meridian<br/>(primary session)"]

    subgraph Packages
        Sources["git sources"] -->|"mars add/sync"| Agents[".agents/"]
        Agents -->|"mars link"| Tool[".claude/ · .cursor/"]
    end

    subgraph Runtime
        Primary -->|"meridian spawn"| Router{"Model router"}
        Router --> Claude["Claude Code"]
        Router --> Codex["Codex CLI"]
        Router --> OpenCode["OpenCode"]
    end

    subgraph State[".meridian/"]
        Spawns["spawns + reports"]
        Sessions["sessions"]
        Work["work items"]
    end

    Agents --> Primary
    Primary --> State
    Claude & Codex & OpenCode -->|"meridian CLI"| State

Agent Packages

meridian-dev-workflow — A dev team: architects, coders, reviewers, testers, researchers, documenters, and the orchestrators that coordinate them.

meridian-base — Core coordination primitives. Included as a dependency of meridian-dev-workflow.

Docs

• Getting Started — prerequisites, harness setup, tool integration
• Commands — full CLI reference
• Configuration — config keys, state layout, environment variables
• MCP Tools — tool surface and payload examples
• Troubleshooting — common issues and diagnostics
• INSTALL.md — agent-friendly install guide

Development

uv sync --extra dev
uv run ruff check .
uv run pytest-llm
uv run --extra dev python -m pyright

See DEVELOPMENT.md for full dev setup.

Contributing

This project is not accepting external contributions at this time. The maintainer is keeping it closed to outside contributions until the codebase is stable enough to support them well. Bug reports and feedback are welcome as GitHub issues.

License

Apache 2.0