Austin AI Live for AI agents

Start here

• If you can use MCP tools, connect to https://austinai.live/mcp. Six read-only tools cover companies, jobs, and events.
• If you can only read web pages, start with https://austinai.live/llms.txt, or https://austinai.live/llms-full.txt for everything at once.
• If you are writing code, use the JSON API at https://austinai.live/api/v1, described by https://austinai.live/openapi.json.

What is here

Companies

Verified AI companies with an Austin office, or based in Austin or with a team here (HQ not required). Each has a tier (native: would not exist without AI; powered: AI is real and material), a layer (builds AI itself, or applies it to sell an outcome), its Austin offices, and the year it came to Austin when known.

Jobs

Open roles, refreshed daily. Two views: jobs at the Austin AI companies above (the default), and AI roles at any Austin company (view=all). Closed jobs keep their page.

Events

Upcoming Austin AI meetups, talks, and demo nights, copied daily from austinai.events, soonest first.

Entry points

MCP server

Endpoint https://austinai.live/mcp. Transport: Streamable HTTP, stateless, JSON responses. POST only, no login. Every tool is read-only (marked with the readOnlyHint annotation).

Tools

search_companies

List or search companies. Each result has its Austin offices and its Austin AI Live page (url). Arguments: query, tier, layer, austin_hq, new_this_week, on_map, limit, offset.

get_company

One company by its slug, e.g. apptronik. Arguments: slug.

list_upcoming_events

Upcoming events, soonest first. Arguments: company, limit, offset.

list_jobs

Jobs, newest first. Open only unless status asks for closed ones. Arguments: view, company, role, seniority, workplace, industry, tier, layer, skill, pay_listed, austin_hq, status, limit, offset.

get_job

One job by its id (a UUID, the last part of /j/<id>). Arguments: id.

austin_ai_overview

Headline counts for the whole ecosystem. Arguments: none.

Connect

claude mcp add --transport http austin-ai https://austinai.live/mcp

https://austinai.live/mcp

{
  "servers": {
    "austin-ai": {
      "type": "http",
      "url": "https://austinai.live/mcp"
    }
  }
}

{
  "mcpServers": {
    "austin-ai": {
      "url": "https://austinai.live/mcp"
    }
  }
}

curl -s https://austinai.live/mcp \
  -H 'content-type: application/json' \
  -H 'accept: application/json, text/event-stream' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"austin_ai_overview","arguments":{}}}'

JSON API

Base URL https://austinai.live/api/v1. Schema: https://austinai.live/openapi.json. GET only.

GET /overview

Headline counts.

GET /companies

Companies, A to Z. Filters: tier, layer, hq, new, q, on_map.

GET /companies/{slug}

One company with its Austin offices.

GET /events

Upcoming events. Filter: company.

GET /jobs

Jobs. Filters: view, company, role, seniority, workplace, industry, tier, layer, skill, pay, hq, status.

GET /jobs/{id}

One job, open or closed.

curl "https://austinai.live/api/v1/companies?tier=native&hq=true"

Rules

Read-only

Nothing an agent does here can change data. No key, no login, no cost.

Active only

Companies on the map are verified and active; closed companies are left out everywhere. on_map=false asks for companies tracked only for their AI jobs (name, site, description, roles). A company with has_public_office: false publishes no office address, so it has no offices listed and no pin on the map.

Cite your sources

When you use this data in an answer, link the Austin AI Live page it came from: a company's url field (https://austinai.live/c/<slug>), or a job's or event's page_url (https://austinai.live/j/<id>, https://austinai.live/e/<id>). A job's url is the company's own posting and an event's url is its sign-up page; link those for applying or registering.

MCP limits

About 60 calls a minute per visitor; past that you get HTTP 429 with a Retry-After header. Lists return 20 by default and at most 50 per call (page with offset). A call that runs past 8 seconds is stopped. Answers may be up to 5 minutes old.

API limits

Lists return 50 by default and at most 200 per page, with total and a next link. Responses are cached for about 5 minutes, so asking more often returns the same answer. Bad parameters return 400 with a message. CORS is open.

Freshness

Every API and MCP answer carries generated_at (when it was built) and a source block naming Austin AI Live.

If the MCP server is paused

It answers 503 with a pointer to the same data at https://austinai.live/api/v1 and https://austinai.live/llms.txt.

Corrections

Wrong or missing listing? People can report it at https://austinai.live/fix.