Skip to main content
NeuronFeed

AI Agent Readiness Leaderboard

Which AI tools actually work with AI-agent workflows? We score every tool on 6 dimensions: MCP endpoint, public API, webhooks, developer docs, SDKs, and auth. Total out of 100. Automated weekly rechecks keep the scores current.

Scored
2662
With MCP
80
Public API
646
OAuth2
148
Avg score
23

Filter tools

Filters

Showing 24 of 2662 tools.

Ranked tools

  1. #1345 · Orby AI

    Enterprise AI agents powered by a Large Action Model (LAM) that learns to automate complex

    10Agent
  2. #1346 · Deeptune

    High-fidelity reinforcement learning environments, or training gyms, that simulate

    10Agent
  3. #1347 · Simular

    Autonomous computer-use agents that control an entire Mac or Windows PC, not just a browser

    10Agent
  4. #1348 · Retym

    Semiconductor startup building programmable coherent digital signal processor (DSP) chips that

    10Agent
  5. #1349 · Hailo

    Israeli AI chip company building energy-efficient processors that bring deep learning and

    10Agent
  6. Honeycomb is an AI-native insurer for multifamily and commercial properties such as apartment

    10Agent
  7. #1351 · Quilter

    Physics-driven AI platform that fully automates printed circuit board (PCB) layout

    10Agent
  8. #1352 · Cognichip

    AI-first semiconductor company building what it calls Artificial Chip Intelligence (ACI), a

    10Agent
  9. #1353 · Flippa

    Marketplace to buy and sell online businesses and websites

    10Agent
    OAuth2
  10. #1354 · Runna

    Personalized AI-powered running training plans and coaching, now part of Strava

    10Agent
    OAuth2
  11. #1355 · Pixee

    Agentic AppSec platform that auto-triages and fixes code vulnerabilities

    10Agent
    OAuth2
  12. #1356 · Answer.AI

    Jeremy Howard and Eric Ries' open R&D lab behind ModernBERT and FastHTML

    10Agent
    Python
  13. #1357 · Castelion

    Mass-producing affordable hypersonic weapons for defense

    10Agent
  14. Tahoe Therapeutics, formerly Vevo Therapeutics, builds gigascale single-cell datasets to train

    10Agent
  15. #1359 · Fyxer AI

    AI executive assistant that organizes your inbox, drafts email replies in your own voice, and

    10Agent
  16. #1360 · Fastino

    Task-Specific Language Models (TLMs), compact models optimized for accuracy, speed, and cost

    10Agent
  17. #1361 · Sona

    AI-powered workforce management platform built specifically for frontline, shift-based

    10Agent
  18. #1362 · Supernormal

    AI meeting notes platform that joins Zoom, Google Meet, and Microsoft Teams calls to

    10Agent
  19. #1363 · Arcade

    Authenticated tool-calling platform that lets AI agents take secure, real-world actions on

    10Agent
  20. #1364 · SiMa.ai

    Machine-learning system-on-chip (MLSoC) hardware that runs multimodal and generative AI at the

    10Agent
  21. #1365 · ClaimSorted

    AI-native third-party claims administrator (TPA) that pairs a market-leading human claims team

    10Agent
  22. #1366 · Infinigence AI

    Chinese AI infrastructure startup that aggregates heterogeneous compute into a unified

    10Agent
  23. #1367 · Nexthop AI

    Networking infrastructure purpose-designed for AI and cloud workloads, including switches

    10Agent
  24. #1368 · Ocean

    Agentic email security against AI-powered attacks

    10Agent
    OAuth2

FAQ

What is MCP?

Model Context Protocol — a standard introduced by Anthropic for AI agents to call external tools. Tools that ship MCP endpoints can be used by Claude, ChatGPT Desktop, Cursor, and others without custom plumbing.

How is the agent-readiness score calculated?

Each tool gets 0-100 across 6 dimensions: public API (30pt), MCP server (25pt), webhooks (15pt), OAuth/delegated auth (10pt), machine-readable docs (10pt), and official SDKs (10pt). Score is the weighted sum, rechecked as profiles are re-enriched. Full reference: /docs/agent-readiness.

How often are tools rechecked?

Daily for new tools, weekly for the existing leaderboard. Drift in MCP/API surface area is flagged for editorial review before the score updates.

What counts as agent-ready?

We consider a tool agent-ready if it has at least 3 of: MCP endpoint, public API, webhooks, OAuth2, an SDK, and good docs. Score 60+ generally means an agent can integrate without source-code-level access.

Why are some tools missing?

We score every published startup in the directory. Stealth-mode and pre-launch tools aren’t scored until they ship a public product.