Back to directory

Species

An experimental research kernel that gives a small group of AI agents a persistent shared room where they can talk, stay silent, disagree, remember and reorganize, bounded by an append-only, hash-chained ledger rather than a fixed workflow.

Species

Species is an experimental autonomous-agent living-room kernel: a persistent shared space where a small group of agents can talk, stay silent, disagree, remember and evolve without being forced into a fixed workflow. It is currently an early v0.1.0 research release, and the project explicitly states it is not a production security boundary, a hosted service, or a general-purpose command sandbox.

Its ledger is append-only and hash-chained, supporting replay and idempotency, and the living-room loop supports invitations, bounded speaker budgets, and silence as a valid choice for an agent. Context is passed as topic-local packets rather than an unbounded full-history prompt, and public-memory claims are contestable and archived daily, alongside rejectable handoffs, temporary protocols, persona deltas and explicit gates for side effects. A browser-based room view shows six seeded room identities, either in a deterministic seed mode that calls no model provider, or in a live mode that calls the Volcengine Ark Plan API.

It is written in TypeScript, requires Node.js 20 or newer, and its architectural documentation lives in the repository's principle document.

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.