Back to directory

LISA

A self-hosted personal AI agent with persistent identity, self-authored goals, scheduled autonomous activity, and a private journal, reachable over Telegram, Discord, Slack, Feishu, iMessage, and webhooks.

LISA

LISA is a self-hosted personal AI agent built on a streaming agent loop with a plugin and hook system, sandboxed shell access, subagents, and session resume, connecting to Anthropic, OpenAI, Gemini, and about 20 further OpenAI-compatible providers, or to local models through Ollama or LM Studio. Beyond a system prompt, each installation writes its own identity, purpose, and values once at creation, sets its own goals that drive scheduled autonomous activity through cron or launchd, and periodically reflects on its own state, journaling, adjusting its skills, and leaving a note about what happened while the user was away.

It is reachable over Telegram, Discord, Slack, Feishu, iMessage, and webhooks, keeps a read-only mailbox over IMAP or Gmail OAuth, and maintains a self-tended, wiki-style knowledge base built from links and subscribed feeds. Its coding work can be routed through a separately connected Claude Code, Codex, or Copilot installation using the credentials that installation already has. The code is open source, distributed as a TypeScript project.

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.