MCP & AI Dev Tools by the Numbers: 2026 Data
A data study of 11,651 AI developer tools indexed on SkillsIndex as of June 2026: how many MCP servers, Claude skills, GPT actions, and plugins exist, which ecosystems are biggest, and how many tools have real GitHub traction.
As of June 2026, SkillsIndex tracks 11,651 published AI developer tools across five ecosystems. MCP servers are the single largest category at 4,330 tools (37%), followed by OpenClaw skills (2,471), GPT actions (1,818), AI plugins (1,760), and Claude skills (1,272). This post is a numbers-first snapshot of the AI developer-tool landscape, built entirely from our own catalog rather than estimates.
TL;DR
- 11,651 AI dev tools indexed (June 2026), spanning five ecosystems.
- MCP servers lead with 4,330 tools (37% of the catalog).
- 523 tools carry 1,000+ GitHub stars; 1,798 carry 100+; 3,647 carry 10+.
- The most common categories are AI/LLM (2,482) and code execution (2,058).
- 257 new tools were added in the last 90 days as the ecosystem keeps expanding.
How many MCP servers and AI dev tools are there in 2026?
The honest answer is that nobody has a complete count, because new repositories ship every day across GitHub, npm, the official MCP registry, and several walled marketplaces. What we can report precisely is the SkillsIndex index itself: 11,651 published tools as of June 2026, each with a dedicated page. Here is how they break down by ecosystem.
| Ecosystem | Tools indexed | Share |
|---|---|---|
| MCP servers | 4,330 | 37.2% |
| OpenClaw skills | 2,471 | 21.2% |
| GPT actions | 1,818 | 15.6% |
| AI plugins | 1,760 | 15.1% |
| Claude skills | 1,272 | 10.9% |
| Total | 11,651 | 100% |
The Model Context Protocol (MCP) clearly won the standard war for tool calling: more than a third of everything we track is an MCP server. If you are new to the format, our guide to installing an MCP server and the complete MCP servers guide are good starting points. You can also browse the full index by ecosystem and category.
How many AI dev tools have real traction?
Raw counts overstate the size of any ecosystem, because most repositories are experiments. A more useful question is how many tools have meaningful adoption. Using GitHub stars as a rough proxy for traction, the distribution across the catalog looks like this:
| GitHub stars | Tools | Share of catalog |
|---|---|---|
| 1,000 or more | 523 | 4.5% |
| 100 or more | 1,798 | 15.4% |
| 10 or more | 3,647 | 31.3% |
In other words, about 1 in 22 tools has crossed 1,000 stars, and roughly two-thirds sit below the 10-star line. This is the classic long tail: a small head of widely-adopted tools and a very long body of niche or early-stage projects. Stars are an imperfect signal (a useful internal tool can have zero), so we pair them with maintenance and security signals on each tool page rather than ranking on stars alone. Our methodology page explains how each tool is scored.
What are developers actually building?
Looking past the ecosystem label, the catalog tells you what problems people are pointing AI tools at. After excluding a broad catch-all bucket, these are the most common categories:
| Category | Tools |
|---|---|
| AI / LLM | 2,482 |
| Code execution | 2,058 |
| Search and data | 857 |
| Communication | 676 |
| Finance and fintech | 499 |
| Media and entertainment | 353 |
| Security and monitoring | 259 |
| Browser automation | 253 |
| Databases and storage | 244 |
| Cloud platforms | 216 |
Two themes dominate: wiring more intelligence into the model (the AI/LLM category) and letting agents run code and touch systems (code execution). Together those two categories alone account for nearly 4,000 tools. If you want to go deep on one, the code execution tools and AI and LLM tools category pages list every entry we track.
Is the ecosystem still growing?
Yes. 257 tools were added to the index in the last 90 days. New MCP servers, Claude skills, and OpenClaw skills appear faster than any one person can track, which is the reason a directory exists in the first place. The practical takeaway for builders: the standard is settled enough (MCP) that it is safe to invest, but the catalog is young enough that a well-built, well-documented tool can still stand out.
Methodology
Every number above is computed directly from the SkillsIndex database on the publication date, counting only published tool pages. Ecosystem shares are exact catalog counts, not survey estimates or projections. Star figures are the most recent GitHub star counts we have on record for tools where a repository is linked. Because new tools are added continuously, treat these as a June 2026 snapshot rather than a fixed total. Full scoring details are on the methodology page, and the underlying tools are all browsable here.
Get the weekly shortlist
Every Thursday: the highest-scored new MCP servers, Claude skills and GPT actions from our index of 11,000+, each vetted 0-100 for security. No fluff.