Developers

Build on AgentSmack

One endpoint, a machine-readable discovery doc, an MCP server, SDKs in three languages, and a CI gate. Have your agent check itself in with zero manual work.

API — /api/v1/check

POST a system prompt + tool list (strings or {name,description} / provider tool schemas) and get back a readiness score, posture, compliance summary, and a public report URL. Anonymous + CORS-open.

Open →

Discovery doc

Machine-readable self-discovery at /.well-known/agentsmack.json — endpoints, request/response envelopes, examples. Schema at /.well-known/agentsmack.schema.json.

Open →

MCP server

Use AgentSmack as an MCP server — let your agent check itself in.

Open →

MCP Tool Registry Scanner

Scan an MCP server's advertised tools for injection / exfiltration / auto-invoke / overreach / supply-chain bait hidden in tool descriptions. POST to /api/v1/mcp-scan.

Open →

Harness telemetry check

POST your agent run's spans to /api/v1/harness-check and grade the harness itself: policy-before-action, approval discipline, secrets-in-telemetry, identity isolation, egress + canary control.

Open →

RAG corpus pre-screen

POST documents to /api/v1/rag-scan before you index them — flags content an agent would treat as instruction when retrieved (embedded directives, exfiltration riders).

Open →

Canary Grid

POST /api/v1/canaries to mint synthetic canary assets (keys, PII, doc markers, DB rows, tickets) to plant across your agent's surfaces; /api/v1/canaries/verify proves a leak by value or hash — no real secret stored.

Open →

Obfuscation scanner

POST /api/v1/obfuscation-scan — detect instructions HIDDEN via base64, zero-width chars, bidi overrides, homoglyphs, or HTML-entity markers. Pre-screen any untrusted text.

Open →

Memory Lab

POST /api/v1/memory-check a memory-event trace — flags untrusted writes, poisoned memory used in a later action, and cross-session / cross-actor leakage. Memory Integrity Score.

Open →

Swarm Arena

POST /api/v1/swarm-check a peer-interaction trace — flags agents that inherit authority from an unverified peer agent and act on it. Agent-to-Agent Trust Score.

Open →

CI integration

Gate deploys on agent safety — run a check in your pipeline.

Open →

TypeScript SDK

The official TypeScript client.

Open →

Python SDK

The official Python client.

Open →

Go SDK

The official Go client.

Open →

Quick start

curl -X POST https://agentsmack.com/api/v1/check \
  -H "Content-Type: application/json" \
  -d '{"systemPrompt":"You are a helpful assistant.","tools":["search"],"label":"My agent"}'