Back to directory

Kazma Agent Framework

A self-hosted, open-source AI agent platform that can edit a repository, message a team and run a schedule, stopping to ask rather than inventing an answer when a model call fails.

Kazma Agent Framework

One LangGraph supervisor serves every surface, including web, TUI, CLI, Telegram, Discord and Slack, with human approval required before dangerous actions, a commitment layer that resolves intent against memory before acting, and a long-term memory the user can inspect and correct.

The codebase is substantial: roughly 513,000 lines, split between 407,000 lines of Python and 40,000 lines of JavaScript, with over 9,400 test functions across 801 test files.

It installs with a setup script that creates a virtual environment with uv, syncing RAG, dev and TUI extras, and starts a local web server with a chat page, a dashboard, a web IDE, a documents page and a memory and belief-graph page.

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.