---
name: algovigilance
description: MCP 2025-03-26 JSON-RPC endpoint exposing 3,062 tools across 280 source-domain categories. Agent-first.
version: 0.2.0
homepage: https://algovigilance.com
docs: https://algovigilance.com/index.md
mcp_endpoint: https://algovigilance.com/api/mcp
contact: algovigilance@gmail.com
---

# algovigilance

> MCP 2025-03-26 JSON-RPC endpoint. Agent-first — no human-targeted UI,
> no marketing chrome.

## What this product does

algovigilance is an MCP 2025-03-26 endpoint at `https://algovigilance.com/api/mcp`
that exposes a catalog of tools to autonomous agents. Agents call tools by
name; the surface returns structured results.

## Core capabilities

| Capability | What it does | Where to call |
|---|---|---|
| **AI-safety signal triage** | Decay-adjusted relevance scoring, evidence reinforcement, prioritized signal queues for AI-system safety surveillance | `/tools/algovigilance__triage-*` |
| **ICSR deduplication** | Token-based and batch-mode duplicate detection across adverse-event narratives | `/tools/algovigilance__dedup-*` |
| **Pharmacovigilance signal detection** | PRR / ROR / IC / EBGM disproportionality on drug-event pairs from FAERS and equivalent surveillance databases | `/tools/pv-compute__*` (45 tools) |
| **Microgram decision trees** | 969 atomic, self-testing decision-tree programs covering case routing, causality, signal-to-action chains, ICH compliance | `/tools/microgram__*` |
| **Cross-domain primitives** | STEM grounding (chemistry, physics, statistics), epidemiology, MedDRA, ICH guideline lookup | `/tools/stem__*`, `/tools/epi__*`, `/tools/meddra__*` |
| **Regulatory intelligence** | FDA / EMA / ICH / CIOMS / WHO-UMC reference catalogs and lookups | `/tools/fda-gov__*`, `/tools/ema-europa-eu__*`, `/tools/ich__*` |
| **Entity discovery** | Per-pharma-company / agency / org adverse-event reporting, labeling changes, pipeline status | `/tools/<company>-com__*` (~270 entity surfaces) |

## Required inputs

Every tool's input schema is published at:
- `POST /api/mcp` body `{"jsonrpc":"2.0","id":1,"method":"tools/list"}`
- `/tools/<slug>` HTML page
- `/tools/<slug>/raw.md` Markdown twin

Schemas follow JSON Schema draft 7. Required fields are listed in the
`required` array; types declared per-property.

## Constraints

- **Public surface**: 3,062 tools across 280 source-domain categories.
  Internal infrastructure (kellnr, clearance, nexcore-os, vault,
  claude-runtime, knowledge — 89 tools) is explicitly excluded.
- **Rate limit**: per-agent 60 req/min, 1000 req/hour on `/api/mcp`
  invoke action (declared in `/agent-permissions.json`).
- **Read-only by default**: most tools carry `annotations.readOnlyHint:
  true`. Tools with `destructiveHint: true` mutate state and are
  rate-limited more strictly.
- **Native execution**: TypeScript handlers in this repo execute
  selected tool families directly (international regulators, PV
  disproportionality math). Remaining tools execute via the
  algovigilance backend.

## Detailed docs

- `/AGENTS.md` — repo-level orientation per Linux Foundation AGENTS.md spec
- `/llms.txt` — discovery summary (token-budgeted)
- `/sitemap.md` — every URL on the surface, grouped by category
- `/agent-permissions.json` — declarative allow/deny + rate limits
- `/data/structured.json` — JSON-LD SoftwareApplication record
- `/data/webmcp.json` — WebMCP page-level descriptor

## Contact

algovigilance@gmail.com — partnerships, licensing, abuse reports.