Make your startup agent-ready

How the score works

NeuronFeed scans every listed company and computes a 0–100 agent-readiness score from six weighted checks (full breakdown in the docs):

Signal	Weight	What we look for
Public API	30	A documented API an agent can call
MCP server	25	A Model Context Protocol endpoint
Webhooks	15	Event push so agents can react, not poll
OAuth	10	Delegated auth an agent can complete
API docs	10	Machine-readable reference (OpenAPI ideal)
SDKs	10	Official client libraries

The 5-item checklist

1. Publish llms.txt

A plain-text file at /llms.txt that tells language models what your product does and where the docs live. Ten minutes of work, disproportionate payoff — it's the first thing agent crawlers request. Ours is at neuronfeed.com/llms.txt as a working example.

2. Decide your AI-crawler policy — explicitly

Your robots.txt should name the AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended, CCBot) and allow or block them deliberately. Silence reads as neglect; some agents treat unknown policy as a no.

3. Ship an OpenAPI spec

Serve it at /openapi.json or /.well-known/openapi.json. Agents can auto-generate working API calls from a spec — it's the difference between "found your product" and "used your product."

4. Stand up an MCP server

MCP is the emerging standard for agent-to-tool connections (Claude, and a growing list of clients, speak it natively). Even a read-only MCP server exposing your core resource earns the 25-point signal and real agent traffic.

5. Add webhooks + OAuth

Webhooks let an agent subscribe instead of poll; OAuth lets a user safely delegate access to one. Together they're another 25 points — and the harder moat, so ship them after the quick wins above.

Measure it

Run the agent-readiness checker on your domain — it performs the same probes our indexer does and shows exactly which checks pass. Then browse the agent-ready leaderboard to see the 100-scorers you're up against.