Akil MCP - New York City Edition

NYC public records for AI agents.

Akil MCP connects Claude, ChatGPT, Codex, and custom agents to a hosted civic data layer built around real workflows, reusable identifiers, and source-backed public records.

The product is the hosted MCP server. This repository is the public connection home for developers and evaluators.

Connect

There is nothing to install for normal use.

MCP endpoint: https://mcp.askakil.ai/mcp

Client	How to connect
Claude	Add Akil as a custom connector, paste the MCP URL, and sign in when prompted.
ChatGPT	Use connectors where available, add the MCP URL, and complete the Akil authorization flow.
Codex	Register Akil as a streamable HTTP MCP server. Developer access can send an API key as a bearer token.
Custom agents	Use OAuth for user-authorized sessions or API keys for backend jobs and partner workflows.

Full setup and docs:

What agents can do

Akil is organized around civic workflows rather than a single dataset.

Evaluate a business location: permits, licenses, inspections, location context
Research a property or building: ownership, deeds, permits, violations, sales, tax records
Follow public money: awards, payments, contracts, agency budgets
Understand influence: lobbying, campaign finance, legislation, votes, hearings
Check an organization: nonprofit registries, audits, funding history, compliance records
Ask about a place: address, BBL, district, neighborhood, radius, polygon, or corridor

Why this exists

A general model can reason. Akil gives it records to reason from.

The useful part is not the number of tools. It is that fragmented public records come back with stable identifiers, source context, and next steps an assistant can actually follow.

Anchor-first design

Akil responses are designed to keep agents oriented around durable civic identifiers:

BBL for tax lots, buildings, ownership, permits, sales, tax records, and violations
EIN for nonprofits, audits, awards, filings, and organization checks
Agency for budgets, contracts, payments, rules, hearings, and payroll context
District or other boundaries for place-based questions
License and application identifiers for regulated businesses
Time window for bounded, usable queries

These anchors let an assistant move from one public record system to another without losing the subject of the question.

Example prompts

text

I am looking at a storefront lease. What public records should I check before signing?

text

Who owns this building, and are there recent permits or violations tied to the tax lot?

text

Show public funding, contracts, and payments tied to this organization.

text

What changed in this council district over the last year?

text

Is this restaurant, liquor, tobacco, cannabis, or sidewalk-cafe location worth checking more closely?

Standards and discovery

For clients, directories, and evaluators that probe machine-readable surfaces:

Server Card: https://mcp.askakil.ai/.well-known/mcp/server-card.json (also at /.well-known/mcp.json)
OAuth resource metadata: https://mcp.askakil.ai/.well-known/oauth-protected-resource (RFC 9728)
Protocol: serves current-era MCP clients (protocol revision 2026-07-28) and 2025-era legacy clients on the same endpoint
Health: https://mcp.askakil.ai/health reports version and live tool count
Tool behavior: every tool is read-only and non-destructive, declares MCP safety annotations, returns a source-attribution footer, and states plainly when a search is capped or empty
Registry manifest: server.json in this repository

Public surfaces

Product: askakil.ai
MCP overview: askakil.ai/mcp
Developer docs: askakil.ai/mcp/docs
Company: akilventures.com
Contact: hello@akilventures.com

About

Akil is built by Akil Ventures, Inc., a Delaware corporation building civic intelligence infrastructure for AI agents.

License

MIT. This public repository documents the hosted MCP server and lightweight connection metadata. The hosted Akil service runs proprietary infrastructure and is not open source.