API & MCP

Developers

Access the Agentic.ai directory via REST API or MCP. No authentication required.

OpenAPI Spec

Machine-readable API definition

Agent Manifest

Capability discovery for AI agents

API Policy

Rate limits, licensing, fair use

Quickstart

The API is read-only and requires no authentication. Rate limit: 100 requests/hour per IP.

# Search for AI tools
curl "https://agentic.ai/api/v1/search?q=coding+agent"

# Get a specific listing with all details
curl "https://agentic.ai/api/v1/listings/cursor-ai"

# Browse open-source tools sorted by agenticness score
curl "https://agentic.ai/api/v1/listings?openSource=true&sort=score"

# Compare two tools side-by-side
curl "https://agentic.ai/api/v1/compare?tools=cursor-ai,claude-code"

# Health check
curl "https://agentic.ai/api/v1/health"

REST API v1

Base URL: https://agentic.ai/api/v1

GET

/categories

List all categories with listing counts

GET

/categories/{slug}

Category detail with its listings

GET

/listings

Browse, filter, and sort listings

GET

/listings/{slug}

Full listing detail with all structured fields

GET

/search?q={query}

Full-text search across listings

GET

/tags

All tags grouped by type

GET

/compare?tools={a},{b}

Side-by-side comparison with verdict

GET

/health

API health and status check

Filter Parameters

Available on /listings:

Param	Values	Description
category	slug	Filter by category
tag	slug	Filter by tag
cohort	PEOPLE, TEAMS	Filter by audience
minScore	0-32	Minimum agenticness score
deployment	CLOUD, SELF_HOSTED, HYBRID, ON_DEVICE	Deployment model
autonomy	COPILOT, SEMI_AUTONOMOUS, FULLY_AUTONOMOUS	Autonomy level
openSource	true, false	Open-source filter
mcpSupport	true, false	MCP support filter
sort	score, newest, name	Sort order (default: score)

Field Semantics

Boolean fields like mcpSupport and openSource use three-state logic:

true — Confirmed yes
false — Confirmed no
null — Unknown / not yet evaluated

Each listing includes a dataCompleteness score (0-1) indicating how many structured fields are populated. Use this to filter out incomplete entries.

MCP Server

Connect AI agents to the directory via Model Context Protocol. 9 tools available.

Claude Code

# Add the Agentic.ai Directory MCP server
claude mcp add --transport http agentic-directory https://agentic.ai/mcp

# Verify it's connected
claude mcp list

# Then in Claude Code, you can ask:
# "Search for open-source coding agents"
# "Compare Cursor vs Claude Code"
# "What's the agenticness score for Devin?"

Other MCP Clients

Any MCP-compatible client can connect using streamable HTTP transport:

Endpoint: https://agentic.ai/mcp
Transport: Streamable HTTP (stateless)
Authentication: None required

Available Tools

MCP

search_listings

Search by query, category, cohort, and structured field filters

MCP

semantic_search

AI-powered natural language search using embeddings

MCP

get_listing

Full listing detail with structured fields and features

MCP

compare_listings

Side-by-side comparison with verdict

MCP

get_agenticness_details

8-dimension agenticness breakdown with evidence

MCP

list_categories

All categories with listing counts

MCP

get_category

Category detail with all its listings

MCP

list_tags

All tags grouped by type

MCP

list_recent

Recently added or updated listings

Code Examples

Python

import requests

BASE = "https://agentic.ai/api/v1"

# Search for coding agents
resp = requests.get(f"{BASE}/search", params={"q": "coding agent"})
for tool in resp.json()["data"]:
    print(f"{tool['name']} — {tool['agenticScore']}/32")

# Compare two tools
resp = requests.get(f"{BASE}/compare", params={"tools": "cursor-ai,claude-code"})
data = resp.json()["data"]
print(f"\nChoose {data['tool1']['name']} if:")
for reason in data["verdict"]["chooseTool1If"]:
    print(f"  + {reason}")
print(f"\nChoose {data['tool2']['name']} if:")
for reason in data["verdict"]["chooseTool2If"]:
    print(f"  + {reason}")

JavaScript / TypeScript

const BASE = "https://agentic.ai/api/v1";

// Get open-source tools with MCP support
const resp = await fetch(`${BASE}/listings?openSource=true&mcpSupport=true`);
const { data, pagination } = await resp.json();

console.log(`Found ${pagination.total} open-source MCP tools:`);
data.forEach(tool => {
  console.log(`  ${tool.name} (${tool.agenticScore}/32) — ${tool.shortDescription}`);
});

Data Exports & Licensing

JSON Dataset

Full directory as JSON with all structured fields.

Download JSON

CSV Dataset

Flat table format for spreadsheets and analysis.

Download CSV

Licensed under CC BY 4.0. Attribution required. See our API Policy for full terms.

Changelog

v1.1.0March 2026

Added structured differentiation fields (deployment, autonomy, MCP support, etc.)
Added /compare endpoint for side-by-side comparisons
Added filter params: deployment, autonomy, openSource, mcpSupport
3 new MCP tools: semantic_search, compare_listings, get_agenticness_details
Added dataCompleteness score to all listing responses
Added /health endpoint
Added X-Request-ID header to all responses

v1.0.0March 2026

Initial public API with 6 endpoints
MCP server with 6 tools
OpenAPI spec, agent.json manifest, llms.txt
JSON and CSV dataset exports

Support

Questions, bugs, or rate limit increases: hello@agentic.ai