AI Security for Agentic Systems
Authenticate agents. Block spoofing. Enforce trust levels.
Website · Documentation · Blog
AI agents are calling each other and invoking tools—but there's no standard way to verify who is calling what. CapiscIO provides the missing identity layer:
- Identity — Ed25519 signatures prove which agent made the request
- Integrity — SHA-256 body hashing detects tampering
- Freshness — 60-second replay windows reject stale traffic
All with sub-millisecond overhead.
# Python SDK for FastAPI/Flask
pip install capiscio-sdk
# MCP server protection
pip install capiscio-mcp
# CLI for validation
npm install -g capiscioThen protect your endpoints:
from fastapi import FastAPI
from capiscio_sdk import SimpleGuard
app = FastAPI()
guard = SimpleGuard(app) # That's it| Repository | Description |
|---|---|
| capiscio-sdk-python | Python SDK with FastAPI/Flask middleware. Drop-in guard for A2A endpoints. |
| capiscio-mcp-python | MCP SDK wrapper. Adds CapiscIO identity and signing to MCP servers. |
| capiscio-core | High-performance Go enforcement engine. Also available as capiscio/guard Docker image. |
| Repository | Description |
|---|---|
| capiscio-node | Node CLI (capiscio). Validate agent cards, test endpoints, run security checks in CI. |
| capiscio-python | Python CLI wrapper for Python-centric environments. |
| validate-a2a | GitHub Action for CI/CD. Validates agent cards and enforces compliance. |
| Repository | Description |
|---|---|
| a2a-demos | Three production-ready agent demos (LangChain, CrewAI, LangGraph) with CapiscIO security. Great starting point. |
| Repository | Description |
|---|---|
| capiscio-rfcs | RFC specifications for trust badges, proof of possession, and protocol extensions. |
CapiscIO addresses 6 of the OWASP Top 10 for Agentic Applications:
| Risk | Coverage |
|---|---|
| AG01: Agent Identity Spoofing | ✅ Direct |
| AG06: Communication Tampering | ✅ Direct |
| AG07: Replay Attacks | ✅ Direct |
| AG10: Audit & Logging Gaps | ✅ Direct |
| AG02: Tool Misuse | ⚡ Partial |
| AG03: Excessive Agency | ⚡ Partial |
Like SSL certificates, but with 5-minute TTLs:
| Level | Validation | Use Case |
|---|---|---|
| 0 | Self-signed | Local development |
| 1 | Domain verified | Production agents |
| 2 | Organization verified | Enterprise |
| 3 | Extended validation | High-trust scenarios |
┌─────────────────────────────────────────────────────────┐
│ CapiscIO Registry (optional) │
│ Agent Discovery · Trust Badges · Events │
└─────────────────────────┬───────────────────────────────┘
│
┌─────────────────┼─────────────────┐
▼ ▼ ▼
┌──────────────┐ ┌──────────────┐ ┌──────────────┐
│ Agent A │ │ Agent B │ │ MCP Server │
│ │ │ │ │ │
│ capiscio-sdk │──▶│ capiscio-sdk │──▶│ capiscio-mcp │
│ (signing) │ │ (verifying) │ │ (verifying) │
└──────────────┘ └──────────────┘ └──────────────┘
Each agent gets a cryptographic identity (DID) and signs requests. Receiving agents verify signatures before processing.
- Try the demos — Clone a2a-demos and run three agents in 5 minutes
- Read the docs — docs.capisc.io
- Open issues — Found an edge case? Let us know
- Join the discussion — Reddit · LinkedIn · X
Open source under Apache 2.0 · Built for the agentic era
