# algovigilance

> Agent-first AI safety MCP surface. 3062 tools across 280 source-domain categories, MCP 2025-03-26 endpoint at /api/mcp, AEO-compliant per Osmani (Google Cloud AI, 2026-04-11). Designed for agent consumption — no marketing chrome, no human-targeted layout. Pharmacovigilance for algorithms.

## Agent entry points

- [/](/index.md): homepage (markdown twin)
- [/tools](/tools.md): tool index — 3062 MCP tools across 280 categories
- [/api/mcp](/api/mcp): MCP 2025-03-26 JSON-RPC endpoint
- [/AGENTS.md](/AGENTS.md): Linux Foundation AGENTS.md spec
- [/agent-permissions.json](/agent-permissions.json): declarative allow/deny + rate limits
- [/sitemap.md](/sitemap.md): every URL grouped by category
- [/data/structured.json](/data/structured.json): JSON-LD SoftwareApplication
- [/data/webmcp.json](/data/webmcp.json): WebMCP page-level advertisement

## Top 20 tool categories (of 280)

All callable via `POST /api/mcp` with method `tools/call`. MCP name shape: `station__<category>__<tool>`. Sample link points at one representative tool per category — POST `tools/list` to enumerate every name with its input schema.

- **microgram** (969 tools) — sample: [list-chains](/tools/microgram__list-chains)
- **pv-compute** (45 tools) — sample: [pv-axioms-domain-dashboard](/tools/pv-compute__pv-axioms-domain-dashboard)
- **body-systems** (40 tools) — sample: [cardio-blood-health](/tools/body-systems__cardio-blood-health)
- **devtools** (40 tools) — sample: [analyze-patterns](/tools/devtools__analyze-patterns)
- **stem** (38 tools) — sample: [stem-bio-behavior-profile](/tools/stem__stem-bio-behavior-profile)
- **observatory** (32 tools) — sample: [viz-ae-overlay](/tools/observatory__viz-ae-overlay)
- **skills** (24 tools) — sample: [skill-categories-compute-intensive](/tools/skills__skill-categories-compute-intensive)
- **molecular-biology** (23 tools) — sample: [chemotaxis-gradient](/tools/molecular-biology__chemotaxis-gradient)
- **science** (22 tools) — sample: [compute-target-score](/tools/science__compute-target-score)
- **vigil** (21 tools) — sample: [vigil-authority-config](/tools/vigil__vigil-authority-config)
- **forge** (20 tools) — sample: [forge-atomize](/tools/forge__forge-atomize)
- **gcloud** (19 tools) — sample: [gcloud-auth-list](/tools/gcloud__gcloud-auth-list)
- **wolfram** (19 tools) — sample: [wolfram-astronomy](/tools/wolfram__wolfram-astronomy)
- **chemistry** (18 tools) — sample: [chemistry-buffer-capacity](/tools/chemistry__chemistry-buffer-capacity)
- **fda** (18 tools) — sample: [fda-assess-risk](/tools/fda__fda-assess-risk)
- **calculate** (17 tools) — sample: [assess-naranjo-causality](/tools/calculate__assess-naranjo-causality)
- **cloud** (17 tools) — sample: [cloud-anomaly-detect](/tools/cloud__cloud-anomaly-detect)
- **education** (17 tools) — sample: [edu-assess](/tools/education__edu-assess)
- **graph-ops** (17 tools) — sample: [aggregate-fold](/tools/graph-ops__aggregate-fold)
- **learning** (17 tools) — sample: [learn-assimilate](/tools/learning__learn-assimilate)

*260 additional categories not shown — see [/tools](/tools) for the complete list.*

## MCP quick-connect

```json
{
  "mcpServers": {
    "algovigilance": { "url": "https://algovigilance.com/api/mcp" }
  }
}
```

## Failure modes monitored

- Drift — production model behavior degradation
- Recursive observer collapse — IAIR / ACA / QPAV governance
- Emergent harm — capabilities below training thresholds

## Compliance

EU AI Act · FDA SaMD · ISO 42001 · NIST AI RMF

## Upstream

Production tool implementations live at https://mcp.nexvigilant.com/mcp (Station MCP server). algovigilance.com is the AI-safety-narrowed front door.

## Contact

algovigilance@gmail.com