Agents, MCP & capabilities · Concept

Build against the contract.

SkillServers specifies one permission model across REST, MCP and the CLI, so an agent cannot do what the operator cannot.

Read the docs Interface matrix

Status Concept Category Agents, MCP & capabilities Surfaces REST · MCP · CLI Contact info@skillservers.com

Quickstart

The shapes below are the specified contract; the host is not live yet.

curl -X POST https://api.skillservers.com/v1/requests \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"scope": "workspace:read", "input": { ... }}'

Specification example. No endpoint at api.skillservers.com is claimed to be running.

Authentication and scope

Tokens carry scope; scope is checked at execution time.

Credentials

Short-lived tokens, per-workspace keys and delegated agent identities. Revocation is immediate.

Scope

Read, propose and execute are separate grants. An agent may propose without being able to execute.

Agent onboarding

What an agent needs before it can act.

Publish an agent card and a machine-readable onboarding document at https://skillservers.com/llms.txt, listing the tools, the scopes each requires and the evidence each returns.

curl https://skillservers.com/llms.txt

SDKs and conformance

Generated from the same description as the API.

SDK

Typed clients

Generated from OpenAPI, pinned to a schema version, with the same error taxonomy as the API.

Checks

Conformance suite

Declared capabilities are tested against the contract before a release is labelled conformant.

Agents

MCP profile

Tools, resources, prompts and tasks mapped to MCP with scope annotations.