Akil MCP - New York City Edition

Give your AI agent NYC records it can understand.

Akil MCP acts as a civic semantic layer for NYC public records, giving AI agents a structured way to investigate funding, property, procurement, lobbying, land use, licensing, housing, and neighborhood context. Start with akil_platform_guide for orientation, then use akil_overview to route specific civic questions.

MCP URLhttps://mcp.askakil.ai/mcp
akil-mcp · session-047
$ akil --connect // authenticated ✓ MCP session opened · civic records available 
231
Source-backed tools
80+
Public-record sources
Guide
Start with akil_platform_guide
Caveats
Source context included
Why Akil exists

Open data is fragmented. Akil gives agents a way through it.

The useful part is not just the number of tools. It is that agents can start from a real-world anchor and receive records with identifiers, interpreted meaning, source context, caveats, and next steps.

01

Official-record retrieval

Agents query public records directly instead of relying on web snippets, memory, or broad search results.

02

Records explained

Fields that are easy to misread come back with plain-English context, so a filing, deadline, payment, status, or empty result is not treated as more than it is.

03

Anchored workflows

Each answer surfaces identifiers and next checks that connect one public system to another: BBL, EIN, agency, license, district, fiscal year.

Trust Checks

Agent-ready means source-aware, not just connected.

Akil's trust checks look for public-record mistakes that agents can easily make: treating a filing as clearance, a name match as identity, a capped page as the full universe, or an empty result as proof that nothing exists.

Sources

Source and caveat checks

Checks whether tool answers keep source names, filters, result caps, freshness notes, caveats, and verification paths visible.

Limits

Coverage and no-result checks

Checks whether windowed, capped, empty, or timed-out responses are framed as bounded source results instead of full clearance.

Roles

Identity and regulatory checks

Checks whether names, owners, vendors, licensees, applicants, permits, payments, and certifications stay separate unless source anchors support a match.

Workflow-aware data

Built for the way people actually ask about the city.

Users don't start with database names. They start with a lease, a building, a vendor, a nonprofit, an agency, a campaign, or a map. Akil translates that into the right public-record path.

Restaurant operator
Open or evaluate a business location
"What permits, licenses, inspections, and location constraints matter at this address?"
ANCHORS
BBLBINCAMISlicense numberapplication ID
TOOLS CALLED
business licensesrestaurant inspectionsliquor licensesOCM cannabis recordsDOB permits
Pulling DOB permit history → 18 records since 2019. Latest: ALT-2 filed 2024-03-14. CAMIS active. Liquor license expires 2025-09.
The anchor layer

Records connect through stable threads.

Akil is designed so an assistant can move from one dataset to the next without losing the subject of the question.

AKIL
ANCHORS
BBLProperty thread
EINOrganization thread
AgencyGovernment thread
DistrictPlace thread
LicenseOperating thread
TimeChange thread
Hover a node to follow its thread →
Agent-ready records

A response your assistant can actually use.

Akil does not just hand an agent a row of data. It keeps the source value, plain-English meaning, caveats, identifiers, and next checks together so the answer stays useful after the first reply.

WHAT THE AGENT RECEIVES
QuestionCan this storefront support the use my client has in mind?
Anchor

BBL 1012970023

Records checked

DOB filings, certificate-of-occupancy records, violations, PLUTO, ACRIS, and nearby license records.

Meaning

Separates a filing from an approval, a permit from a signoff, and a missing row from proof that nothing exists.

Caveat

Not legal clearance. Confirm final status with the agency, lease counsel, or the right licensed professional.

Next checks

Check CO/legal use, open violations, recent filings, and license-radius constraints.

example packetsource + meaning + caveat
{
  "type": "akil.record_packet.v1",
  "anchor": {
    "kind": "property",
    "bbl": "1012970023"
  },
  "finding": {
    "source": "NYC DOB",
    "sourceValue": "Application filed",
    "meaning": "A filing exists. It is not the same as an approved permit or final signoff."
  },
  "caveat": "Verify current status before relying on this for a lease or closing.",
  "nextChecks": [
    "certificate of occupancy",
    "open violations",
    "license fit"
  ]
}
What agents can ask

Natural-language questions, public-record answers.

Examples of good starting points. Each begins with an anchor; the assistant follows the relevant public-record path from there.

01 / 10Restaurant operatorAddress + license
"I am looking at this address for a cafe. What permits, licenses, inspections, and nearby public records should I check before signing?"
Hover to pause · click any segment to jump
Connect

Connect once. Start from a real-world anchor.

Client support varies by AI product. Use the connector interface provided by your client when available; Akil remains one protected remote MCP endpoint.

CONNECTING — CLAUDEIN PROGRESS

Add Akil as a custom connector, paste the MCP URL, and sign in when prompted.

Open Claude settings
Add custom MCP server
3
Paste https://mcp.askakil.ai/mcp
4
Authorize access
One connection · NYC records on demand

Give your assistant access to agent-ready NYC public records.

Start from an address, organization, district, person, program, agency, license, or question.

Read developer docs