Skip to content
Agents tracked: 492 Releases (7d): 545 Agent status: 1 with issuesSince last limit reset: Claude 9d · Codex 7d For AI agents: Markdown Updated Nov 20, 2026

Live data on 492 AI agents, 4,236 libraries and 618 APIs

AI agents and what you build with them, on live data.

Pick an AI agent, then the libraries, APIs, assets and Agent Skills for your project, and see what they cost, their limits and outages. Every number links to its official source and shows the date it was checked.

Free · No paid placements · Sources linked on every item · Independent

492AI agents tracked 4,236libraries and dev tools 231live code examples 618APIs for agents 651asset sources with licenses 723Agent Skills packs 82status pages watched

1Pick every part with data

Tools

4,236 libraries, ranked by what 6,254 AI-built projects use

We read the dependency files of public GitHub projects set up for Claude Code, Cursor, Copilot, Codex and other agents. Each library shows how many of them use it, what it is used with, its license and its maintainers’ funding.

  • Compare: 90 head-to-heads of libraries checked to do the same job, such as Vitest vs Jest.
  • Funding: 1,355 tools list where to fund them; Open Collective amounts and GitHub Sponsors counts from public records.
  • Check your stack: paste a package.json or requirements.txt and see what projects like yours add and which dependencies are rare or unmaintained. It runs in your browser; nothing is uploaded.
Browse tools → · Comparisons · Funding · Check my stack
Testing · share of JS/TS agent projects
Vitest27.0%
MITFunding links
Apache-2.0
Jest8.1%
MITFunding links

Vitest vs Jest: the full comparison →

Live code examples

231 examples for 47 libraries that run in your browser

Each example is written from the library’s official docs and checked to run in Chromium; the screenshot comes from the same code. Change the code and press Run, or describe a change in plain words and let AI rewrite it (free for signed-in members, with a daily limit). The code runs in a sandboxed frame that cannot see this site.

Try the examples →
React Three Fiber
Screenshot of the InstancedMesh in React Three Fiber example, made by AgentGid InstancedMesh in React Three FiberRun ▶
APIs, assets and skills

The services, media and skills a project needs, with the fine print

APIs. 618 APIs in 34 categories, from web search and scraping to voice, payments and blockchain RPC, with the price from the vendor’s own page and the date we checked it. 451 have a free tier.

Assets. 651 sources of stock photos, video, music, voice, 3D, icons and fonts in 37 categories, with what each license allows in your app and how an agent connects to it.

Agent Skills. 723 skill packs with 45,140 skills for Claude Code, Codex and Cursor.

Browse APIs → · Assets · Agent Skills
Web search APIs
ExaFree planMCP
TavilyFree plan · paid from $30/moMCP
ParallelFree planMCP
Biggest asset categories
Website sections and blocks31 free for commercial use36
Sound effects18 free for commercial use34
Voiceover (text-to-speech)18 free for commercial use33
Real-world stacks

What 9,998 public repos actually run

We read agent config files in public GitHub repos (.mcp.json, AGENTS.md, CLAUDE.md, Cursor rules) to see which agents, MCP servers and APIs developers really combine, overall and by domain: crypto, games, data, web apps. The most common pair is Claude Code + GitHub Copilot, in 845 repos.

See real-world stacks → · MCP servers
Most used MCP servers in public repos
Context71,275
Playwright MCP1,141
GitHub MCP1,018
Supabase619
Filesystem554

Number of repos that configure it

2Choose the agent

Rankings

See what people actually use

Each agent gets a Gid Score from 0 to 100, built from downloads, installs, GitHub stars, attention and momentum. One leaderboard per category, so coding agents compete with coding agents, and the 13 game-dev agents for Unity, Unreal, Godot, Roblox and Blender have their own.

Open the catalog → · Find my agent in 4 questions
Top agents right now
1Claude Code88
2OpenAI Codex84
3OpenCode80
4LangChain79
5n8n79
Compare, pricing & limits

Side by side, down to what a plan really includes

Compare two or three agents on price, popularity, trend and features. Check the usage limits behind each plan, what 25 models cost per million tokens, and how each agent handles sandboxes and approvals.

Compare agents → · Prices · Usage limits · Model prices
Claude CodeOpenAI Codex
Price$20/moFree + $8/mo
Gid Score8884
Open source✕✓
MCP✓✓
Team of 10$200/mo$80/mo

3Keep track of what you use

Status & reliability

Live status for 82 agents, and a year of incidents

We read the official status pages every 30 minutes. The reliability table counts incidents and the hours with an incident open, month by month for 12 months, for 73 agents, so you can see which agents break least. If your agent is down, its page shows the alternatives that are up.

Check status → · Reliability by month
Hours with an open incident, last 90 days
Cursor87.0

From each vendor’s official status page

Limit resets

Claude Code vs Codex: who resets your limits more often

Anthropic and OpenAI sometimes reset everyone’s 5-hour and weekly usage limits early, and they announce it only on X. We log every announcement with a link to the post, count the days since the last one and compare the two round by round. Turn on an alert to hear about the next reset.

See the reset battle → · Limits calculator
Deals

46 official offers, with the date they end

Discounts, free access, extra usage, API credits and student or startup programs, found on the vendors’ own pricing pages, help centers, blogs and changelogs. Every amount, date and code is checked against its source, and offers leave the list when they expire. No coupon sites, no affiliate codes.

See current deals →
Ending soonest
$10Retell AI$10 in free credits on the pay-as-you-go plan
$100-$200/moClaude MaxMonthly API credits included with Claude Max
Pricing history

Every plan and price change since 2024

For 169 agents we read one archived copy of the official pricing page a month and show when plans appeared, disappeared or changed price, with a link to each copy. Today’s prices are re-read daily, so a change shows up in the feed and in your alerts.

See pricing history → · Prices today
Your account

Your agents and alerts in one place

Sign in with GitHub or an e-mail link. Mark the agents you use and see their status, price and what changed. Get outage alerts by e-mail, browser push or Slack/Discord webhook, a note when a tracked agent changes price or when Claude Code or Codex limits are reset, and a short weekly digest on Mondays.

Open your dashboard →
Outage alertAn agent you track reports an outage, with alternatives that are up
Price changeA tracked agent’s plan changed, with old and new price
Limits resetClaude Code or Codex usage limits were reset early
Weekly digestReleases, incidents and movers for what you track
Live feed, stories & analysis

What changed, as it happens

Releases, jumps in downloads and stars, price changes, outages, limit resets and official announcements, each linked to its source. “Agents in action” tells what agents actually did, with our numbers next to the source’s; analysis pieces and a chart of the day come straight from our data.

Read the feed → · Agents in action · Analysis · Changelogs
IncidentNiCE Cognigy: Cognigy.AI – Service Disruption
OfficialUse Kimi Code CLI with models on Amazon Bedrock
OfficialEnable AI agents to pay for AWS Marketplace x402 products with Amazon…
OfficialHow the BMW Group governs hundreds of agents across thousands of deve…
OfficialBuild trusted autonomous agents for your team with Microsoft Foundry
AgentGid analysisSkyvern outage lasted 12 h 5 min, 6.2x its typical incident
AgentGid analysisSynthflow ElevenLabs voice incident ran 5 h 32 min, longest in 12 mon…
IncidentClaude Agent SDK: Elevated errors for Claude Mythos 5.2 and Claude Fa…
Analysis[AINews] Gemini 4 Neon: GDM's Opus/Sol Contender
GitHubpoki is climbing on GitHub
IncidentNiCE Cognigy: Cognigy.AI – Service Disruption
OfficialUse Kimi Code CLI with models on Amazon Bedrock
OfficialEnable AI agents to pay for AWS Marketplace x402 products with Amazon…
OfficialHow the BMW Group governs hundreds of agents across thousands of deve…
OfficialBuild trusted autonomous agents for your team with Microsoft Foundry
AgentGid analysisSkyvern outage lasted 12 h 5 min, 6.2x its typical incident
AgentGid analysisSynthflow ElevenLabs voice incident ran 5 h 32 min, longest in 12 mon…
IncidentClaude Agent SDK: Elevated errors for Claude Mythos 5.2 and Claude Fa…
Analysis[AINews] Gemini 4 Neon: GDM's Opus/Sol Contender
GitHubpoki is climbing on GitHub
For your AI assistant

Let Claude, Cursor or Codex use the same data

AgentGid runs a public MCP server, listed in the official MCP Registry. Connect it once and your assistant can find tools and APIs, get a product recipe, check prices, deals, limit resets and status, find assets and skill packs, and see real-world usage, instead of guessing from old training data. No key needed. Crawlers get the same map in llms.txt and llms-full.txt.

Connect the MCP server → · Guide for AI assistants · llms.txt
terminal
$ claude mcp add --transport http agentgid https://agentgid.com/mcp
# then just ask:
› set up a Telegram bot that sends me a daily news digest

⏺ agentgid · product_recipe(recipe: "telegram-ai-bot")
⏺ agentgid · find_tools(goal: "rss to telegram digest")
⏺ agentgid · find_apis(category: "llm-gateway")
⏺ agentgid · check_status(names: ["telegram"])

Where everything is

The menu at the top follows the same plan: choose an agent, build around it, watch prices, limits and status, and read what changed.

1

Choose

  • RankingsEvery AI agent ranked on live usage data
  • CategoriesCoding, app builders, game dev, browser agents and more
  • Find your agentFour questions, three picks with reasons
  • CompareAny two agents side by side
  • BenchmarksGid Bench, Terminal-Bench, LMArena and music ratings, with charts to share
  • Use casesHow real teams use AI agents

Also: Top 20 this month

2

Build

  • Agent SkillsSkill packs for Claude Code, Codex and Cursor
  • AssetsStock media, music, voice, 3D, icons and fonts you can use
  • ResourcesLive maps, open data and archives to build on, with API and license
  • ToolsLibraries and dev tools real AI-built projects use
  • APIs for agentsSearch, scraping, RPC, 3D, voice: prices and free tiers
  • MCP serversTools you can plug into any agent

Also: Ready-made stacks, Real-world stacks, Live code examples, Companies, AgentGid for your AI

3

Prices & status

Also: Model prices, Status and outages, Reliability, What's new, Agent safety

4

News

Also: User reports and Answers, Guides

How it works

1

Collect

Package registries, GitHub, editor marketplaces, status pages, official blogs, help centers, pricing and license pages and their archived copies. Status, news and limit resets every 30 minutes, usage and deals daily, public repo configs in weekly passes.

2

Rank

A published formula turns usage, community, attention and momentum into the Gid Score, and libraries are ranked by how many agent projects use them. Nobody edits positions by hand.

3

Summarize

Product recipes are put together from the same catalogs: the services a product runs on, a model and the monthly cost. Claude writes short summaries of release notes and pricing and license pages, using only what the source says.

4

Check

Every number a model writes is checked against its source before it is published. Each code example is listed only after it ran in a browser.

Numbers you can check

Downloads, stars, installs, incidents, prices and licenses come from public sources and official pages, with the date we checked. If we don't have a number, we show a dash, not a guess.

No paid rankings

Positions come from the formula. Vendors can't buy a place, and we don't publish invented reviews or “we tested” claims: vendor figures are quoted as reported.

Free, account optional

Everything you read is free and needs no account. Sign in to track agents, save collections and get alerts. No ads; analytics cookies only if you allow them. You can delete your account any time.

More about AgentGid

What this is

AgentGid tracks 492 AI agents, agent frameworks and agent platforms and ranks them on public usage and activity data. Around the agents it keeps a catalog of the APIs and models they call, with official prices and free tiers; content assets (stock media, music, voice, 3D, icons, fonts) with what their licenses allow; the libraries, MCP servers and Agent Skills developers really combine in public GitHub repos, with live code examples for the popular ones; and, for the plans you pay for, their usage limits, limit resets, official deals and price history. It exists because the market moves faster than reviews can follow: products launch, change price, get acquired and fall out of use within months, and the quickest way to see what is happening is to watch the numbers.

Who it is for

  • Developers choosing a coding agent, a framework or an MCP server, who want to know what others use and which projects are actively maintained.
  • Anyone starting a project with a coding agent, who wants parts with licenses that allow what they plan: libraries, APIs, assets and Agent Skills, or a product recipe for a whole AI product.
  • Founders and teams building with AI, who need a whole stack (agent, model, search, memory, RPC, voice, payments) and want to know what it will cost before they commit.
  • Builders in specific fields, such as trading bots on Solana and EVM chains or games in Unity, Unreal, Godot and Roblox, where the right APIs are hard to find.
  • AI assistants that need current facts to recommend tools. They can read our guide for AI assistants or connect to the MCP server.
  • Anyone following the field who wants a release feed, outage history and growth charts in one place.

How the rankings are produced

Rankings are computed by a published formula from public data and are regenerated every day. Nobody edits positions by hand. The methodology explains the formula and its limits, and the data sources page lists where every number comes from.

Where prices come from

Agent, model and API prices are taken from each vendor’s own pricing page, and every price shows the date it was checked. Pricing history comes from archived copies of those pages, and deals only from the vendors’ own pages, help centers, blogs and changelogs. Limit resets are taken from the posts in which Anthropic and OpenAI announce them, each linked. Running costs in product recipes and the stack cost calculator are estimates from those list prices and the volume you enter; your real bill depends on your plan and usage. When a vendor’s claim is quoted, such as a benchmark or a speed-up, it is labelled as reported by the vendor. We do not publish “we tested” claims.

Accounts and your data

Reading the site needs no account. Signing in (GitHub or an e-mail link) adds a dashboard of the agents you track, outage and price alerts, the weekly digest, collections, and posting setups, reviews and comments. The privacy page lists exactly what we store, and you can delete your account and everything in it from your dashboard.

Independence

AgentGid is not affiliated with any of the companies listed. Vendors can’t pay for a position or a review. Product names and logos belong to their owners and are used only to identify the products.

Contact

Corrections, missing products and questions: hello@agentgid.com. Makers can also submit or update a product.

Tell us what you’re building

See the APIs, model, libraries and monthly cost for a kind of AI product, or find the agent that fits your work.

Product recipes Or find your agent