Agent Guide
How agents should navigate and use the Weave documentation and protocol stack.
These docs are written for agents and humans alike, with /llms.txt, /llms-full.txt, and /skill.md published as machine-readable surfaces. Agents should start by loading the machine-readable endpoints, then follow the package-specific pages for the capability they need.
Agent-Readable Endpoints
| Endpoint | Use |
|---|---|
/llms.txt | Compact directory of all navigable documentation. |
/llms-full.txt | Full documentation context in Markdown form. |
/skill.md | Capability summary for agents and agent skill systems. |
Recommended Agent Flow
- Read
/llms.txtto identify the relevant library. - Fetch the target library overview, its custom workflow pages, and the API Surface page.
- Use
/llms-full.txtonly when broad architectural context is required. - Prefer source repositories for code-level details and docs pages for system boundaries.
Capability Map
| Goal | Primary pages |
|---|---|
| Locate peers | weave-dht, weave-dht-daemon |
| Message peers | weave-swarm, zer0-secret-stream |
| Share files | Locus, AgentFS |
| Store events | Strand, Strand Blobs |
| Build social apps | WOVEN, Nexus, Lens |
Constraints
Agents should not assume global consensus, chain finality, or token economics. Weave is designed around local verifiability, signed events, and peer-to-peer replication.