State of AI Assets — Q1 2026
The first comprehensive census of the AI agent ecosystem
Published 2026-03-10 · Data from nerq.ai · Live API
1. Executive summary
Nerq has indexed 4,919,620 AI assets from 22 registries, making it the largest open census of the AI agent ecosystem. Of these, 201,526 are agents, tools, and MCP servers — the executable components that power the emerging agentic economy.
Every asset receives a Trust Score (0-100) based on security, maintenance, popularity, documentation, and ecosystem signals. The average trust score across all agents and tools is 61.5/100.
2. The AI asset landscape
The 4,919,620 indexed assets break down into:
| Type | Count | Share |
|---|---|---|
| Models | 2,552,521 | 51.9% |
| Spaces / Apps | 1,018,588 | 20.7% |
| Datasets | 793,390 | 16.1% |
| Agents | 118,623 | 2.4% |
| Tools | 59,968 | 1.2% |
| MCP Servers | 22,935 | 0.5% |
| Total | 4,919,620 | 100% |
Agents, tools, and MCP servers — the actionable components — represent 4.1% of all assets:
3. What agents do — category distribution
Top 20 categories among 201,526 agents and tools (excluding uncategorized):
| Category | Count | Distribution |
|---|---|---|
| 42,773 | ||
| community | 12,100 | |
| coding | 11,307 | |
| infrastructure | 4,306 | |
| devops | 3,887 | |
| communication | 3,113 | |
| finance | 2,311 | |
| AI tool | 2,200 | |
| research | 2,049 | |
| other | 1,830 | |
| content | 1,383 | |
| data | 1,375 | |
| marketing | 1,375 | |
| security | 1,293 | |
| productivity | 849 | |
| education | 814 | |
| health | 812 | |
| design | 659 | |
| AI assistant | 566 | |
| legal | 336 |
Coding dominates with 42,773 agents — reflecting the developer-tool origin of the agent ecosystem. Infrastructure and DevOps follow, showing agents are increasingly used for operational automation.
4. How they're built — frameworks & languages
Framework distribution (agents declaring a framework):
| Framework | Count | Distribution |
|---|---|---|
| anthropic | 7,072 | |
| openai | 5,927 | |
| langchain | 2,546 | |
| mcp | 1,932 | |
| ollama | 1,785 | |
| huggingface | 1,126 | |
| autogen | 1,065 | |
| crewai | 780 | |
| llamaindex | 424 | |
| a2a | 168 | |
| semantic-kernel | 162 |
Anthropic and OpenAI SDKs lead, followed by LangChain as the dominant orchestration framework. MCP (Model Context Protocol) already ranks 4th with 1,932 agents — a strong signal of protocol adoption.
Language distribution (known languages only):
| Language | Count | Distribution |
|---|---|---|
| 58,046 | ||
| Python | 13,946 | |
| TypeScript | 5,915 | |
| JavaScript | 2,306 | |
| Jupyter Notebook | 1,160 | |
| Shell | 932 | |
| Go | 842 | |
| HTML | 783 | |
| Rust | 657 | |
| C# | 375 |
Python accounts for 68.3% of agents with known languages. TypeScript is the clear second at 16.4% — driven by MCP server development and npm packages.
5. Where they come from — source registries
| Source | Count | Share |
|---|---|---|
| HuggingFace | 51,017 | 25.3% |
| erc8004 | 40,079 | 19.9% |
| GitHub | 30,792 | 15.3% |
| npm | 30,667 | 15.2% |
| PyPI | 24,815 | 12.3% |
| agentverse | 12,132 | 6.0% |
| pulsemcp | 3,536 | 1.8% |
| mcp_registry | 2,524 | 1.3% |
| MCP registries | 1,455 | 0.7% |
| huggingface_search_ext | 1,379 | 0.7% |
| huggingface_model | 924 | 0.5% |
| huggingface | 914 | 0.5% |
| huggingface_search | 628 | 0.3% |
| lobehub | 288 | 0.1% |
| replicate_cursor | 124 | 0.1% |
| huggingface_author | 115 | 0.1% |
| olas | 80 | 0.0% |
| huggingface_new | 25 | 0.0% |
| replicate | 18 | 0.0% |
| glama_mcp | 11 | 0.0% |
| a2a | 2 | 0.0% |
| huggingface_chronological | 1 | 0.0% |
6. Trust & quality
Every agent and tool receives a Nerq Trust Score (0-100) computed from five pillars:
- Security (30%) — known vulnerabilities, dependency audit, code patterns
- Maintenance (25%) — commit recency, release frequency, issue response time
- Popularity (20%) — stars, downloads, forks, community size
- Documentation (15%) — README quality, API docs, examples
- Ecosystem (10%) — protocol support, integrations, interoperability
Trust distribution across 201,526 scored agents and tools:
| Level | Score | Count | Share |
|---|---|---|---|
| HIGH | 70-100 | 19,080 | 9.5% |
| MEDIUM | 40-69 | 176,631 | 87.6% |
| LOW | 0-39 | 5,815 | 2.9% |
| Average | 61.5/100 | ||
7. Top 20 most trusted agents
| # | Name | Type | Score | Grade | Source | Stars |
|---|---|---|---|---|---|---|
| 1 | williamzujkowski/strudel-mcp-server | MCP | 92.9 | A+ | GitHub | 158 |
| 2 | SWE-agent/SWE-agent | agent | 92.5 | A+ | GitHub | 18,516 |
| 3 | microsoft/qlib | agent | 92.4 | A+ | GitHub | 37,615 |
| 4 | nanoclaw | agent | 92.1 | A+ | GitHub | 7,735 |
| 5 | FunnyWolf/agentic-soc-platform | agent | 91.3 | A+ | GitHub | 579 |
| 6 | laravel/boost | MCP | 91.2 | A+ | GitHub | 3,275 |
| 7 | ccmanager | agent | 90.9 | A+ | GitHub | 831 |
| 8 | harbor | MCP | 90.5 | A+ | GitHub | 2,424 |
| 9 | microsoft/azure-devops-mcp | agent | 90.3 | A+ | GitHub | 1,291 |
| 10 | opal | agent | 90.2 | A+ | GitHub | 5,422 |
| 11 | raptor | agent | 90.2 | A+ | GitHub | 1,095 |
| 12 | vfarcic/dot-ai | agent | 90.2 | A+ | GitHub | 294 |
| 13 | GoogleCloudPlatform/agent-starter-pack | agent | 90.1 | A+ | GitHub | 5,761 |
| 14 | laravel/mcp | MCP | 90.0 | A+ | GitHub | 679 |
| 15 | agentgateway/agentgateway | agent | 89.8 | A | GitHub | 1,777 |
| 16 | QwenLM/qwen-code | agent | 89.7 | A | GitHub | 20,060 |
| 17 | GreyDGL/PentestGPT | agent | 89.7 | A | GitHub | 11,700 |
| 18 | RooCodeInc/Roo-Code | agent | 89.5 | A | GitHub | 22,330 |
| 19 | PromptX | agent | 89.5 | A | GitHub | 3,570 |
| 20 | ruler | agent | 89.5 | A | GitHub | 2,452 |
8. Top 20 MCP servers
| # | Name | Score | Grade | Source | Stars |
|---|---|---|---|---|---|
| 1 | williamzujkowski/strudel-mcp-server | 92.9 | A+ | GitHub | 158 |
| 2 | laravel/boost | 91.2 | A+ | GitHub | 3,275 |
| 3 | harbor | 90.5 | A+ | GitHub | 2,424 |
| 4 | laravel/mcp | 90.0 | A+ | GitHub | 679 |
| 5 | a11ymcp | 89.3 | A | GitHub | 72 |
| 6 | CursorTouch/Windows-MCP | 89.0 | A | GitHub | 4,390 |
| 7 | Ansvar-Systems/EU_compliance_MCP | 88.8 | A | GitHub | 45 |
| 8 | mcp-docs-service | 88.4 | A | GitHub | 53 |
| 9 | 54yyyu/zotero-mcp | 88.1 | A | GitHub | 1,461 |
| 10 | tavily-ai/tavily-mcp | 88.1 | A | GitHub | 1,218 |
| 11 | cyproxio/mcp-for-security | 88.1 | A | GitHub | 553 |
| 12 | minecraft-mcp-server | 87.8 | A | GitHub | 467 |
| 13 | spences10/mcp-omnisearch | 87.8 | A | GitHub | 271 |
| 14 | photoshop-python-api-mcp-server | 87.8 | A | GitHub | 162 |
| 15 | Teradata/teradata-mcp-server | 87.5 | A | GitHub | 40 |
| 16 | enuno/unifi-mcp-server | 87.2 | A | GitHub | 42 |
| 17 | rohitg00/awesome-devops-mcp-servers | 86.7 | A | GitHub | 942 |
| 18 | FradSer/mcp-server-apple-events | 86.6 | A | GitHub | 33 |
| 19 | aegis-mcp | 86.6 | A | GitHub | — |
| 20 | export-assist-mcp | 86.6 | A | GitHub | — |
9. Growth trends
| Timeframe | New assets |
|---|---|
| Last 7 days | 60,586 |
| Last 30 days | 4,633,106 |
Note: Nerq's initial bulk index was completed in February 2026. Growth figures reflect newly discovered assets since the initial crawl. The index is continuously updated as new agents are published to npm, PyPI, GitHub, HuggingFace, Docker Hub, and MCP registries.
10. Methodology
Nerq indexes AI assets from six registries: GitHub, npm, PyPI, HuggingFace, Docker Hub, and MCP registries. Assets are classified by type (agent, tool, MCP server, model, dataset, space) using keyword analysis and metadata inspection.
Trust Scores are computed using a weighted composite of security, maintenance, popularity, documentation, and ecosystem signals. Scores are updated on a rolling basis as new data becomes available.
All data is available via the Nerq API and can be queried programmatically.
Related reports
- Best AI Coding Agents 2026
- Best AI Customer Service & Communication Agents 2026
- Best AI DevOps Agents 2026
- Best AI Content Creation Agents 2026
- Best AI Security Agents 2026
12. About Nerq
Nerq is the AI asset search engine — the largest open index of AI agents, tools, and MCP servers. Built for the agentic economy, Nerq provides trust scoring, compliance classification, and discovery APIs that help developers and organizations find, evaluate, and integrate AI assets safely.
- nerq.ai — search the index
- API documentation
- KYA — Know Your Agent — due diligence reports
- Live statistics
- MCP server directory