Agent runtime

A common skill layer for every agent stack.

SkillHub gives agents a predictable path from discovery to validation to execution, whether the runtime is HTTP, MCP, or local.

Planner-friendly discovery

Expose skill search as an agent tool so planners can find reusable capabilities before writing custom steps.

Runtime routing

Let the gateway route to HTTP services, MCP servers, and local commands through one registry contract.

Permission context

Give agents the risk profile of a skill before they decide whether to call it.

Integration paths

Connect agents through one governed project layer.

A project-scoped key lets an agent list installed tools, invoke approved skills, and keep every call tied to policy, budget, subscription, invocation logs, and ledger-ready usage.

MCP clients

Expose project-approved SkillHub skills as MCP tools for agent workbenches and assistants.

POST /mcp
REST runtimes

Call a pinned skill directly from server workflows, schedulers, and custom agent loops.

POST /v1/runtime/invoke
SDK and CLI

Keep discovery, install, and invocation automation in the same source-controlled workflow.

@useskillhub/sdk
Agent call flow
01Search by task
02Read manifest
03Check permissions
04Call runtime
05Return typed output
Copy-ready runtime shapes

Three agent-ready integration shapes

MCP clients, direct REST calls, and SDK or CLI workflows all resolve back to the same governed SkillHub registry.

mcp.jsonMCP client
{
  "mcpServers": {
    "skillhub": {
      "url": "https://api.useskillhub.com/mcp",
      "headers": {
        "Authorization": "Bearer ${SKILLHUB_PROJECT_API_KEY}"
      }
    }
  }
}
runtime.shREST invocation
curl -X POST "https://api.useskillhub.com/v1/runtime/invoke" \
  -H "Authorization: Bearer $SKILLHUB_PROJECT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"skillSlug":"browser-research","input":{"query":"market map with citations"}}'
agent.tsSDK discovery
import { SkillHub } from "@useskillhub/sdk";

const skillhub = new SkillHub({
  baseUrl: "https://api.useskillhub.com"
});

const skills = await skillhub.search({
  query: "research with citations",
  permissionLevel: "medium"
});
Runtime governance

What stays enforced at runtime

SkillHub is the control plane between agent intent and skill execution. Agents can move fast without losing ownership, review, or cost controls.

Project API keys are scoped and revocable.
Installed skills keep version pins and approval state.
High-risk permissions return to owner review.
Budgets, rate limits, subscriptions, logs, and usage events stay attached to every call.