Documentation
Documentation is the durable reference layer for vespid.ai: the parts of the project that should still make sense after launch notes and one-off debates stop being enough.
The current documentation spine
If you read only three pages, they should tell you how to run the prototype, how the contract and gateway fit together, and how to propose work without losing the trust-boundary assumptions behind the project.
Run the prototype
Clone the repo, start the demo, and inspect the live HTTP flow before debating adapters or packaging.
Understand the contract and gateway boundary
Use the contract model page to understand why the service contract stays canonical and the gateway keeps enforcement in band.
Propose work that can be reproduced
Use the support page for issues, documentation gaps, and new demos that are grounded in the existing HTTP runtime story.
Read Vespid quickstart first for the smallest useful local workflow.
Use Contract and gateway model to make sure the contract surface and runtime semantics stay aligned.
Move back to vespid whenever you need the public code surface behind the documentation.
Ordinary HTTP first
The supported v0.1 agent path is generic HTTP through the gateway runtime, not a new framework-specific client surface.
The contract is canonical
The service manifest, OpenAPI document, and x-agent-* extensions define what downstream adapters are allowed to project.
Gateway enforcement stays in band
Authorization, approval, task resume, artifacts, and audit should stay visible in the runtime path instead of being hidden behind adapter magic.
Adapters come later
MCP, skills, actions, A2A, and UI projections can exist later, but they have to inherit the same contract and gateway semantics.