Progi is open source and self-hosted — Star us on GitHub

Open source · Self-hosted

MCP-native workflow engine for
any AI harness Cursor OpenCode Claude Code Zed GitHub Copilot Antigravity

Progi teaches your agent how YOU like to get things done. So you can do your best work without re-explaining your process or losing context between sessions.

Start MCP Server View on GitHub

Demo GIF - board moving on its own

"No human touched that board."

How Progi works

1

Describe your workflow

Describe your process in plain language. You can be detailed or just provide a rough idea. Progi stores it as a structured workflow with per-step playbooks.

2

Run tasks, stay in the loop

“Hey Progi, continue working on xy task.” Your agent loads the workflow, works through each step using your playbooks, and loops you in at critical checkpoints to review output.

3

Monitor progress

Progi Monitoring gives you a live view of every running and completed task - status, progress, and the full output history across all your workflows.

4

Optimize as you go

Tweak playbooks between runs. Because workflows live in a database and survive context resets, every future task picks up your changes automatically - your process gets sharper with each iteration.

Components

Progi MCP Server

A Model Context Protocol server that plugs into any MCP-compatible AI harness. It lets you create and refine workflows, start tasks, and make your agent follow your playbooks - looping you in at critical checkpoints along the way.

  • ✓  Create and refine workflows
  • ✓  Start and advance tasks
  • ✓  Agent follows your playbooks
  • ✓  Human-in-the-loop checkpoints

Progi Monitoring

A local web app that gives you full visibility into your ongoing and completed work. Browse the task board, inspect step-by-step output history, and review or edit your workflow playbooks.

  • ✓  Live task board with status & progress
  • ✓  Full output history per task
  • ✓  Review and edit playbooks
  • ✓  Runs entirely on your machine

Get started

MCP config (VS Code / Cursor / Zed / Claude Code)

{
  "mcpServers": {
    "progi": {
      "command": "uvx",
      "args": ["progi"]
    }
  }
}

Or with Claude Code CLI

claude mcp add progi -- uvx progi

Frequently Asked Questions

How is Progi different from just giving the AI a prompt?
A prompt disappears after one session. Progi makes your workflow persist: it lives in a local database, survives context resets, and produces a record of what was done and what output was generated at each step. You can reuse the same workflow across many tasks - define the process once, then optimize it over time.
Does Progi replace my AI assistant?
No. Progi runs alongside your AI harness as an MCP server. Your AI calls Progi’s tools to read the next step, do the work, and submit output. Progi is the memory and structure layer - your AI does the actual work.
What workflow types are supported?
Sequential, branching/conditional, loop/iterative, human-in-the-loop, and hierarchical (sub-workflows). Progi is generic - any process your AI harness can handle is a good candidate.
Where does my data live?
In a local SQLite file on your machine. Progi doesn’t send any of your data to external services. You control the path with PROGI_DB_PATH.
Is it really free?
Yes - open source, MIT, self-hosted. No paid tier, no usage limits, no telemetry.

Stop re-explaining yourself.

Progi is open source and self-hosted. Drop it in, connect your harness, and your agent finally knows how you work.

Star on GitHub