Skip to main content
AIDiveForge AIDiveForge
Visit Pi Omniagent Extensions

Get This Tool

License: License: unverified
Local-run terms: Clone repository, copy files to ~/.pi/agent/extensions, run npm install, ensure target CLIs on PATH, restart Pi.

Share This Tool

Compare This Tool
📋 Embed this tool on your site

Copy this code to embed a compact tool card:

Pi Omniagent Extensions

FreeOpen SourceSelf-HostedAgentic

Pricing

Model
Free

Summary

Managing three or four coding agent subscriptions means constant context-switching — opening a different terminal, re-explaining the codebase, losing your place in Pi every time you need Codex to handle something Claude Code won't. Pi Omniagent Extensions patches that gap by letting you call Cursor, Codex, Claude Code, and Rovo from Pi's model picker without leaving your workspace.

Each extension is a TypeScript file that wraps one external agent's CLI behind an ACP (Agent Communication Protocol) interface, so Pi treats it as just another selectable model. You pick 'Cursor Sonnet' or 'Opus [claude-code-acp]' from the picker, and Pi routes your turn to that agent running locally in your environment. The architecture is thin by design — four files, an npm install, no hosted API, no backend. That thinness is also the ceiling: this is a single developer's open-source project with two GitHub stars and no stated contributors, so production support expectations need to match that reality. Teams with a single agent workflow get no benefit here.

Bottom line: Pick this if you are already paying for Pi plus at least two other coding agent subscriptions and want to stop switching terminals; skip it if your team needs a maintained, supported integration layer, because a two-star solo project is not that.

Community Performance Report Card

No community ratings yet. Be the first to rate this tool!

Best For: Pi users wanting unified access to several coding CLIs, Developers managing multiple agent subscriptions, Workflows that benefit from ACP-based agent interoperability

Community Benchmarks Community

No community benchmarks yet. Be the first to share a real-world data point.

  • ACP-based agent handoff inside a single Pi session, so you route work to Cursor, Codex, Claude Code, or Rovo without rebuilding workspace context in a separate terminal.
  • All four agent wrappers are in plain TypeScript with no abstraction layer, so reading exactly what the shim does — and patching it when a CLI changes — takes minutes rather than filing a support ticket with a vendor.
  • Fully local execution with no hosted backend, which means no additional data leaving your machine beyond what each agent's own CLI already sends.
  • Free and open-source under no stated proprietary license, so there is no paywall blocking access to any of the four integrations.
  • When any of the four wrapped CLIs updates its interface — Cursor, Codex, Claude Code, or Rovo — the corresponding shim silently breaks or throws until the repo author pushes a fix. A two-star, zero-contributor project has no SLA for that fix. Teams with a dependency on uptime switch to maintaining their own fork immediately.
  • There is no routing logic inside the tool: deciding which agent handles which task is entirely manual, via the Pi model picker. Teams wanting rule-based or cost-optimized routing across agents hit this ceiling on the first attempt to automate distribution and end up writing their own dispatch layer outside the tool.
  • The tool only works inside Pi — it provides zero value to developers not already using Pi as their primary coding agent host, which means any team evaluating a different primary agent surface abandons this entirely rather than adapting it.

Community Reviews

No reviews yet. Be the first to share your experience.

About

API Available
No
Self-Hosted
Yes
Last Updated
2026-06-21T02:16:42.290Z

Best For

Who it's for

  • Pi users wanting unified access to several coding CLIs
  • Developers managing multiple agent subscriptions
  • Workflows that benefit from ACP-based agent interoperability

What it does well

  • Access multiple coding agents from Pi's model picker
  • Maximize AI credits by routing work across Cursor, Codex, Claude, and Rovo
  • Run external agents inside an existing Pi workspace session

Integrations

PiCursorOpenAI CodexClaude CodeAtlassian Rovo

Discussion Community

No discussion yet. Sign in to start the conversation.

Compare Pi Omniagent Extensions

Spotted incorrect or missing data? Join our community of contributors.

Sign Up to Contribute

Community Notes & Tips Community

Be the first to contribute. General notes, observations, gotchas, and tips from people who use this tool day-to-day.

Frequently Asked Questions

Is Pi Omniagent Extensions free?
Yes — Pi Omniagent Extensions is fully free to use. There is no paid tier.
Is Pi Omniagent Extensions open source?
Yes. Pi Omniagent Extensions is open source.
Can I self-host Pi Omniagent Extensions?
Yes. Pi Omniagent Extensions supports self-hosting on your own infrastructure.

Hours Saved & ROI Stories Community

Be the first to contribute. Concrete time/cost savings, with context. e.g. "Cut my code review backlog from 4h to 45m per week."

Pi Omniagent Extensions

Pi Omniagent Extensions is a set of four TypeScript shims — cursor-acp.ts, codex-app-server.ts, claude-code-acp.ts, and rovo-acp.ts — that each wrap one external coding agent’s CLI into an ACP-compatible interface. Once installed via npm, they surface inside Pi’s /model picker as selectable models. You choose ‘GPT-5 [codex-app-server]’ or ‘Cursor Sonnet’ directly inside your existing Pi workspace session, and Pi hands your turn off to that agent’s CLI running locally. No cloud relay, no account linking beyond what each agent already requires.

The differentiating design choice is ACP (Agent Communication Protocol), which provides the common message envelope that lets Pi treat a Cursor process and a Claude Code process as interchangeable model endpoints. That means switching agents mid-session is a picker selection, not a terminal hop. For developers who are already subscribed to several tools and want to route different task types to the agent best suited for them — without rebuilding context each time — this is the mechanical glue that makes that possible.

This fits a narrow but real use case: a single developer or small team sitting on multiple agent subscriptions who wants to maximize credit utilization across those tools from one interface. It breaks down at the boundary of its scope — there is no task routing logic, no fallback handling, and no abstraction above the per-agent CLI. When a CLI it wraps changes its interface, the corresponding shim breaks until the repo is updated. With zero open issues, zero pull requests, and two stars, the update cadence is an open question.

Installation is local only, documented in README.md, HOW_IT_WORKS.md, and USAGE_GUIDE.md in the repository. No API is exposed, no hosted service is involved. Each agent’s own authentication and subscription requirements remain in force — this tool layers on top of what you already have, not a replacement for any of it.