Back to directory

Simplex

A semi-structured specification format, now at version 0.6, for describing work assignments for autonomous agents: a document states required behavior, completion criteria, examples and failure handling, checked by a deterministic Go linter.

Simplex

Simplex is a specification language for describing what an autonomous agent must do without dictating how, so the agent keeps freedom over implementation while the assignment stays checkable. A Simplex document defines one or more functions, each requiring a RULES section stating required behavior, a DONE_WHEN section stating completion criteria, EXAMPLES, and ERRORS describing failure handling; optional elements include stable identifiers and a COVERS section mapping examples back to the rules and criteria they demonstrate.

The repository holds the normative specification itself, four example .simplex documents, and a Go-based linter with a CLI, tests and fixtures that checks a document against the documented rules using a tolerant parser. The authors are explicit that a passing lint result only means those specific checks found no errors - it does not prove a specification is complete, unambiguous or sufficient for any particular agent to execute correctly.

The project's public website and its deployment configuration live in a separate repository; this checkout's own copies of them are legacy and not authoritative.

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.