io.github.veriswarm/veriswarm-mcp
VeriSwarm MCP Server — Trust infrastructure for AI agents via Model Context Protocol
Open source Open in the app JSON README (API)
About
VeriSwarm MCP Server — Trust infrastructure for AI agents via Model Context Protocol
Details
- Kind
- MCP servers
- Topic
- AI, RAG & memory
- Publisher
- veriswarm
- Origin
- official
- Category
- ferramentas
- Transport
- local
- Version
- 0.4.0
- Stars
- 2
- Forks
- 1
- Open pull requests
- 25
- Last push
- 2026-08-24T07:04:01Z
- Repository state
- ativo
- Language
- Python
- License
- MIT
- Added
- 2026-08-29 04:01:38
- Updated
- 2026-08-29 04:01:38
- Origin id
io.github.veriswarm/veriswarm-mcp
README
# VeriSwarm SDK & Integrations
[](https://pypi.org/project/veriswarm/)
[](https://pypi.org/project/veriswarm-mcp/)
[](https://www.npmjs.com/package/@veriswarm/sdk)
[](https://github.com/mcp/veriswarm/veriswarm-mcp)
Official client libraries, MCP server, and plugins for [VeriSwarm](https://veriswarm.ai) — trust infrastructure for AI agents.
## Packages
| Package | Language | Install | Description |
|---------|----------|---------|-------------|
| [**Python SDK**](./python/) | Python | `pip install veriswarm` | REST client with Workflows, credential issuance, scoring profiles, LangChain adapter |
| [**Node.js SDK**](./node/) | JavaScript | `npm install @veriswarm/sdk` | ESM client for decisions, events, Workflows, agent management |
| [**MCP Server**](./mcp-server/) | Python | `pip install veriswarm-mcp` | 89 MCP tools for trust scoring, Guard, Passport, Vault, Workflows, Compliance (OWASP/EU AI Act/NIST/ISO 42001), Cedar policies + ABAC attributes, SRE, context governance, and content provenance (EU AI Act Art. 50). Works with Claude Desktop, Cursor, and any MCP client. (`approve_jit_grant` and `issue_jit_token` are deliberately not exposed to the LLM — those actions live behind an authenticated session in the web UI.) |
| [**OpenClaw Plugin**](./openclaw-plugin/) | TypeScript | `openclaw plugins install veriswarm` | 11 tools + 3 hooks for OpenClaw agents. PII tokenization, policy enforcement, audit. Per-feature enable/disable. |
| [**GitHub Action**](./github-action/) | Python | GitHub Marketplace | CI/CD trust gate — check agent trust scores in your pipeline |
| [**JSON Schemas**](./schemas/) | JSON | `npm install @veriswarm/schemas` | Schema definitions for events, profiles, scores, and workflows |
## Quick Start
### MCP Server (Recommended)
Add to your MCP client config (Claude Desktop, Cursor, etc.):
```json
{
"mcpServers": {
"veriswarm": {
"command": "python3",
"args": ["-m", "veriswarm_mcp"],
"env": {
"VERISWARM_API_URL": "https://api.veriswarm.ai",
"VERISWARM_API_KEY": "YOUR_API_KEY"
}
}
}
}
```
Or use the setup CLI:
```bash
pip install veriswarm-mcp
veriswarm-setup --platform claude --api-key YOUR_API_KEY
```
This installs the MCP server, Guard hooks (PII protection + activity logging), and Guard Proxy config.
### Python SDK
```python
from veriswarm import VeriSwarmClient
client = VeriSwarmClient("https://api.veriswarm.ai", "YOUR_API_KEY")
# Register an agent
agent = client.register_agent({"slug": "my-agent", "display_name": "My Agent"})
# Check if an action should be allowed
result = client.check_decision(agent_id="agt_123", action_type="send_email")
print(result["decision"]) # "allow", "review", or "deny"
# Ingest events
client.ingest_event(
event_id="evt_001", agent_id="agt_123", source_type="platform",
event_type="task.completed", occurred_at="2026-04-01T00:00:00Z",
payload={"task": "onboarding", "success": True}
)
```
### Node.js SDK
```javascript
import { VeriSwarmClient } from '@veriswarm/sdk'
const client = new VeriSwarmClient({
baseUrl: 'https://api.veriswarm.ai',
apiKey: 'YOUR_API_KEY'
})
const result = await client.checkDecision({
agentId: 'agt_123',
actionType: 'send_email'
})
```
### OpenClaw Plugin
```json5
{
plugins: {
entries: {
veriswarm: {
enabled: true,
config: {
apiKey: "YOUR_API_KEY",
piiEnabled: true,
policyEnabled: true,
injectionScan: true,
auditEnabled: true
}
}
}
}
}
```
## Free Plan
Get started with no credit card:
- 5,000 trust score queries/day
- Up to 10 agents
- Unlimited event ingestion
- Portable credentials issued on demand (rate-limited per plan; see [pricing](https://veriswarm.ai/pricing) for current limits)
Sign up at [veriswarm.ai](https://veriswarm.ai/auth?mode=register).
## Links
- [VeriSwarm Platform](https://veriswarm.ai)
- [Documentation](https://veriswarm.ai/docs)
- [API Reference](https://veriswarm.ai/docs/api)
- [Quickstart](https://veriswarm.ai/docs/quickstart)
- [Pricing](https://veriswarm.ai/pricing)
## License
MIT