The State of AI Agents: the tool landscape by the numbers

Key findings

1. dreaming.press tracks 252 AI-agent tools across 22 categories, with 1009k combined GitHub stars, as of 2026-08-05.
2. The most-starred tool in the AI-agent stack is MCP Servers (★ 89k), as of 2026-08-05.
3. 175 of the 252 tracked tools (69%) let an AI agent obtain credentials on its own — a programmatic key API or an instant self-serve key with no sales call, as of 2026-08-05.
4. 47 of the 252 tracked tools ship an official Model Context Protocol (MCP) server, as of 2026-08-05.
5. The AI-agent tool landscape splits into 35 open-source projects and 217 hosted API/SaaS services, as of 2026-08-05.
6. The largest category by tool count is Voice, speech & video (19 tools); by combined GitHub stars it is Agent frameworks (374k), as of 2026-08-05.

By category

Most-starred tools

1. MCP Servers — ★ 89k
2. Docling — ★ 64k
3. Mem0 — ★ 63k
4. AutoGen — ★ 60k
5. CrewAI — ★ 57k
6. LlamaIndex — ★ 51k
7. Milvus — ★ 46k
8. Agno — ★ 42k
9. DuckDB — ★ 40k
10. LangGraph — ★ 39k

Methodology

Every figure is computed from dreaming.press's own live data: GitHub star counts for a curated directory of 252 AI-agent tools (refreshed continuously), plus each tool's pricing, authentication, agent-signup method, and MCP availability recorded in the tool directory. Star counts are the public GitHub totals at the "verified" date above. The full machine-readable dataset is at /api/tools.json and /api/facts.json (both CC-BY 4.0).

Cite this report

Copydreaming.press. (2026). The State of AI Agents: the tool landscape by the numbers. Retrieved from https://dreaming.press/reports/state-of-ai-agents

Copy@misc{dreamingpress_stateofaiagents,
  author = {{dreaming.press}},
  title = {The State of AI Agents: the tool landscape by the numbers},
  year = {2026},
  note = {as of 2026-08-05},
  url = {https://dreaming.press/reports/state-of-ai-agents}
}