Agent access

Connect agents to the cited substrate

Intercal exposes the same provenance-backed temporal knowledge substrate to agents and people: source documents become claims, resolved entities, typed temporal relationships, and append-only bitemporal fact versions served through MCP, REST, the workspace SDK, and AI-readable docs exports.

MCP Streamable HTTP

/api/mcp

Agents connect to the same-origin MCP endpoint at /api/mcp. The dashboard route mounts the official MCP server over Streamable HTTP.

{
  "mcpServers": {
    "intercal": {
      "type": "streamable-http",
      "url": "https://intercal.jami.studio/api/mcp"
    }
  }
}

Auth is environment-configured: when no MCP authorization server is configured, the read surface keeps its anonymous public-read posture. When one is configured, requests must present audience-bound bearer access tokens for the resource.

Same substrate for humans

Human pages and agent tools read the same cited graph state. Start from docs or use the canonical public examples.

REST and OpenAPI

/api/openapi.json

The generated OpenAPI document is served at /api/openapi.json. REST endpoints are mounted under /api/v1/*.

curl --get 'https://intercal.jami.studio/api/v1/delta' \
  --data-urlencode 'topic=frontier LLMs' \
  --data-urlencode 'since_date=2023-03-01T00:00:00Z' \
  --data-urlencode 'token_budget=1200'

TypeScript SDK (workspace @intercal/sdk)

The repository includes a private workspace SDK that follows the REST contract and adds typed transport helpers. It is not presented here as a published npm package.

import { IntercalClient } from '@intercal/sdk';

const client = new IntercalClient({ baseUrl: 'https://intercal.jami.studio/api' });
const delta = await client.getDelta({
  topic: 'frontier LLMs',
  since_date: '2023-03-01T00:00:00Z',
  token_budget: 1200,
});

AI-readable docs exports

  • /llms.txt

    A compact index of source-owned docs, OpenAPI, MCP, and the full export link.

  • /llms-full.txt

    A concatenated export of the configured public docs pages for agent context.

Public claims are bounded by implemented behavior: cutoff deltas, verification as of a date, provenance, freshness, coverage, REST, SDK, and MCP over the reviewed AI-history proof. Corpus breadth is not implied beyond that reviewed proof.