Skip to main content

Multi-Agent Collaboration CLI with flock-based locking and event-driven workflow

Project description

Agent Pulse

Multi-Agent Collaboration CLI — flock-based coordination, event-driven workflow, zero overhead.

Pulse is a command-line tool that coordinates multiple AI agents working on a shared project. It provides atomic locking, queue-based issue routing, per-agent heartbeat timeouts, and an automated daily summary loop — all backed by a simple file-based store.


Features

  • Atomic flock locks — race-condition-free claim/complete cycles via POSIX fcntl
  • Queue-based routing — issues route to the right agent with configurable SLA per queue
  • Heartbeat-aware guardian — stale locks auto-release after heartbeat × 1.5 minutes
  • Event-driven daily summary — ordered turn-taking with idempotent trigger/collect
  • Async mirror.pulse/ internal store syncs to async/ for human visibility
  • Configurable via JSON — all rules externalized to config.json; no code changes needed
  • Dual output format--format json for machine consumption, text for humans
  • Dry-run mode--dry-run previews every mutation without side effects

Installation

pip install mai-cli

Or install from source:

git clone https://github.com/yourname/mai-cli.git
cd mai-cli
pip install -e .

Quick Start

Step 1 — Initialize a project

pulse project init MyProject

Step 2 — Create an issue

pulse --project MyProject issue new programmer-questions "How should we handle input buffering?"

Step 3 — Have the assigned agent claim it

pulse --project MyProject issue claim REQ-001

Step 4 — Mark it done

pulse --project MyProject issue complete REQ-001 "Decision: ring buffer, 60Hz polling"

Step 5 — Inspect queues

pulse --project MyProject queue check

Architecture

.pulse/                        async/
├── queues/<queue>/  Issue files  <queue>/  Human-visible mirror
├── locks/           flock files  (internal only)
├── processing/      Active issues  <queue>/
├── decisions/       Conclusion logs  decisions/
├── history/         Audit logs  history/
├── events/          Daily-summary triggers  (internal only)
└── config.json      All collaboration rules

Lock protocol: When an agent claims an issue, Pulse acquires an flock(2)-based file lock. The lock expires after heartbeat × 1.5 minutes if the agent fails to heartbeat. The lock guardian cron command automatically releases stale locks.

Daily summary flow: trigger → each agent write in order → collect merges all summaries into a single report.


Command Reference

Issue

pulse issue new <queue> <title> [--ref REQ-XXX]
pulse issue amend <issue-id> <remark>
pulse issue claim <issue-id>
pulse issue complete <issue-id> <conclusion>
pulse issue list [queue]
pulse issue show <issue-id>
pulse issue escalate <issue-id>

Queue

pulse queue check [queue] [--overdue]
pulse queue blockers

Lock

pulse lock check <issue-id>
pulse lock force-release <issue-id>
pulse lock guardian

Log

pulse log history [--date YYYY-MM-DD] [--agent NAME]
pulse log write <agent> <type> <summary> [status]

Daily Summary

pulse daily-summary trigger
pulse daily-summary write <agent> <content...>
pulse daily-summary collect

Escalation

pulse escalation gen <issue-id>

Bitable

pulse bitable sync-status
pulse bitable retry

Project

pulse project init <project-name>

Configuration

Edit .pulse/config.json in your project root:

{
  "queues": {
    "programmer-questions": {
      "handler": "designer",
      "sla_minutes": 120,
      "id_prefix": "REQ"
    }
  },
  "agents": {
    "programmer": { "heartbeat_minutes": 17 },
    "designer":   { "heartbeat_minutes": 29 }
  },
  "daily_summary_order": ["programmer", "designer", "architect", "narrative", "techartist"]
}

Legacy field compatibility

ownerhandler, sla_hourssla_minutes are automatically converted.


Supported Platforms

  • OS: Linux / macOS (POSIX required for flock)
  • Python: 3.8, 3.9, 3.10, 3.11, 3.12

License

MIT License

MIT License — see LICENSE for the full text.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

mai_cli-1.1.0.tar.gz (21.3 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

mai_cli-1.1.0-py3-none-any.whl (25.9 kB view details)

Uploaded Python 3

File details

Details for the file mai_cli-1.1.0.tar.gz.

File metadata

  • Download URL: mai_cli-1.1.0.tar.gz
  • Upload date:
  • Size: 21.3 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.12.3

File hashes

Hashes for mai_cli-1.1.0.tar.gz
Algorithm Hash digest
SHA256 a02dabf14f2e3e1a3df661b7d0bb361de903605395c87bbf10e52ce991841bc2
MD5 c93f1e5d7fe813afa9d939aa50d17667
BLAKE2b-256 197383e6be06a3726d0847859f20eb75046ddaf9bc28392a14525e01765c9d3c

See more details on using hashes here.

File details

Details for the file mai_cli-1.1.0-py3-none-any.whl.

File metadata

  • Download URL: mai_cli-1.1.0-py3-none-any.whl
  • Upload date:
  • Size: 25.9 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.12.3

File hashes

Hashes for mai_cli-1.1.0-py3-none-any.whl
Algorithm Hash digest
SHA256 6af6de969c5f8219c3c70da6fe88ce8122658af25f244548d7fc8e44b43d0a0b
MD5 89282d8debdfc63356c1cba435f57076
BLAKE2b-256 6d75720455d688eb234d490b304aeb27bcd625cd65fe3bd8f4b8ba42318d1ec7

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page