Arachne documentation

Welcome to the Arachne documentation. Arachne is a DSPy-native runtime for turning goals into typed, inspectable, self-healing execution graphs.

The docs follow the Diátaxis structure so readers can choose the right material for the job: tutorials for learning, guides for doing, explanation for understanding, and reference for exact details.

Start here

graph TD
    A[New reader]
    B[Getting started]
    C[Architecture]
    D[CLI reference]
    E[Developer guide]
    A --> B
    A --> C
    A --> D
    A --> E

Reader goal	Recommended page
Install Arachne and run a first goal	Getting started
Understand the runtime lifecycle	Architecture
Learn command syntax	CLI reference
Work on the codebase	Developer guide
Understand DSPy-native design	DSPy-native concepts
Understand large-output handling	Pointer pattern

Documentation map

graph LR
    A[Tutorials]
    B[Guides]
    C[Explanation]
    D[Reference]
    A --> B
    B --> C
    C --> D

Tutorials

• Getting started — set up Arachne and run a first agent graph.

Guides

• Developer guide — local setup, workflow, and project conventions.
• Testing guide — running tests and understanding pytest configuration.
• Creating skills — adding reusable expert protocols.
• MCP setup — connecting Model Context Protocol servers.

Explanation

• Architecture — system design, graph lifecycle, data flow, and trust boundaries.
• Architecture overview — high-level runtime view.
• Graph orchestration — topology weaving and execution.
• MCP integration — tool integration design.
• Self-healing — AutoHealer and recovery strategies.

Concepts

• DSPy-native — why Arachne uses DSPy signatures over prompt chains.
• Triangulated evaluation — rules, semantic scoring, and human escalation.
• Pointer pattern — large-output spillover and downstream retrieval.

Reference

• CLI reference — command syntax, options, and examples.
• Schema reference — Pydantic models and topology contracts.
• Coding standards — style, linting, and review expectations.
• Security Policy — policy and disclosure process.

Project resources

• README — public project landing page.
• AGENTS.md — AI coding assistant context.
• ROADMAP.md — project vision, phases, and milestones.