Back to directory

Pinvou Agent

An open-source desktop AI agent workspace that combines everyday task work, prompt-to-editable visual design, and bringing an external coding agent such as Codex or Claude Code into a real project, all in one interface.

Pinvou Agent

Pinvou Agent is a desktop workspace organized around three ways of working. For everyday tasks, it combines attachments, a local knowledge base, specialist personas, Skills, MCP tools and workflows to research, analyze, write and deliver reusable files rather than just chat replies. For design work, a natural-language prompt produces an editable visual, such as a poster or data visualization, that can then be adjusted directly by selecting elements or by describing further changes. For coding, it brings Codex, Claude Code or Kimi into the same workspace over the Agent Client Protocol to read and edit a real project, run commands, and surface plans, tool steps and permission requests.

Every file the agent creates or edits collects automatically in an artifact panel, editable Markdown artifacts can be revised directly or by asking the agent, and a local knowledge base supports full-text and vector retrieval across attached collections. It can run fully offline against a local model or connect to any OpenAI-compatible endpoint.

It is written in Rust and released under the MIT license.

More in Agent frameworks

See all 902
  • Recurse

    recurse.run

    A serverless harness that lets a coding agent design, test, and deploy its own specialist sub-agents as tools, MCP servers, or bots, defined by a schema-validated manifest.

  • ProofForge

    shulgin.is-a.dev

    An AI-agent pipeline that decomposes a mathematics problem, proves each piece, and formalizes the result in Lean 4 and Mathlib until the Lean kernel accepts it, with six of its proofs already merged into Google DeepMind's formal-conjectures repository.

  • local-coder

    github.com

    Builds a local coding-agent team on top of Ollama and OpenCode by detecting the machine's hardware, recommending models for seven specialist roles, and testing whether each assigned model can actually use tools, edit a repository and pass verification.