What is io.github.bolyra/mcp?

MCP auth middleware: gate tool calls so only authorized agents can act.

How do I install io.github.bolyra/mcp?

Clone io.github.bolyra/mcp from https://github.com/bolyra/bolyra and configure it as an MCP server in your client (Claude Desktop, Cursor, or VS Code).

Is io.github.bolyra/mcp open source?

Yes, io.github.bolyra/mcp is open source under the Apache-2.0 license. Source code is available at https://github.com/bolyra/bolyra.

What category does io.github.bolyra/mcp belong to?

io.github.bolyra/mcp is listed under Security & Auth MCP servers on MCPFind.

io.github.bolyra/mcp — MCP Server for Security & Auth

Bolyra

Verified agent actions: authorization policy, signed receipts, and tamper-evident audit for AI agent tool calls.

Domain: bolyra.ai
Company: ZKProva Inc.
License: Apache-2.0 (with DCO sign-off — every commit requires Signed-off-by:)

What this is

When an AI agent calls a tool, Bolyra proves who — and what — authorized the action. Put the gateway in front of any MCP server (or embed the verifier in your agent platform) and every tool call gets four checks: credential verification, per-tool policy, replay protection, and a signed audit receipt.

Bolyra ships in two tiers on the same verifier contract:

Bolyra Core — classical crypto, no circuits, no trusted setup: per-tool policy, nonce replay protection, ES256K-signed action receipts, credential binding against registered credentials, and JWT-based delegation claims (@bolyra/delegation). This is the gateway's --dev mode; with a credentials section (or --credentials) configured, claimed identities and permission masks are enforced against the registry — unknown, forged, or expired credentials are denied fail-closed with signed receipts. Without registered credentials, permission claims remain self-asserted (the gateway warns at startup and flags the receipts).
Bolyra ZK — the privacy upgrade: humans prove uniqueness via a Semaphore v4-style enrollment circuit; AI agents prove EdDSA-signed credentials with cumulative-bit permissions; a delegation circuit narrows scope one-way (permissions can only drop, never widen), with the delegation path hidden from the verifier. A handshake binds the human and agent Groth16 proofs to a shared session nonce, verified atomically on-chain.

Core gets you policy-gated, replay-protected, receipted actions. ZK gets you cryptographically bound verified actions without disclosure — the verifier learns that the predicate holds, not your credentials, policies, or delegation graph. You don't need ZK to gate your first MCP server with one command.

Building an agent platform? Bolyra plugs in as an external verifier and gives you an enterprise security capability — verified agent actions — without rebuilding auth. See bolyra.ai.

Repository layout

circuits/        Circom 2 circuits + snarkjs/rapidsnark proving
contracts/       Hardhat — Solidity verifiers + on-chain registry
sdk/             @bolyra/sdk (TypeScript, public API)
sdk-python/      bolyra (Python — pure types + subprocess bridge)
integrations/    langchain, crewai, mcp, openclaw, payment-protocols
spec/            DID method, IETF-style draft, conformance runner
examples/        mcp-demo, provider-mock
docs/            quickstart, OWASP agentic mapping

Quickstart

See sdk/QUICKSTART.md for the TypeScript SDK quickstart.

Build & test

npm install
npm run compile:circuits
npm run compile:contracts
npm test                              # circuits fast + contracts
FULL_PROOF=1 npm run test:circuits:slow  # full Groth16/PLONK proving (~2 min)

Protocol Conformance

112 executable test vectors verify the implementation matches the protocol specification. CI runs them on every PR. See spec/CONFORMANCE.md for the current generated report.

npm run conformance          # run all 112 vectors
npm run conformance:report   # generate spec/CONFORMANCE.md

Implementing your own EVC host? 28 of those vectors test host behavior and run against any host in any language, with no install:

npx @bolyra/evc-conformance --host "/path/to/your-host --flags"

See spec/IMPLEMENTER.md for the full pass path — the Host-Under-Test contract, failure classes, troubleshooting, and a CI snippet.

Contributing

This project requires a Developer Certificate of Origin (DCO) sign-off on every commit. Use git commit -s. See CONTRIBUTING.md for details.

License

Apache-2.0. See LICENSE.

Set up in your AI client

AI clientOperating system

Merge this template into ~/Library/Application Support/Claude/claude_desktop_config.json. Keep existing servers. Add any arguments, credentials, and permissions required by the maintainer; this template has not been install-tested.

json

{
  "mcpServers": {
    "io-github-bolyra-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@bolyra/mcp"
      ]
    }
  }
}

Restart Claude Desktop completely for changes to take effect. Confirm the server appears connected in the client’s tool list, then try a read-only example from its documentation.

Claude Desktop setup reference

Compatible MCP Clients

io.github.bolyra/mcp works with any MCP-compatible client. Copy the config snippet from the Configuration section above and add it to the file shown for your client, then restart the application.

Claude Desktop~/Library/Application Support/Claude/claude_desktop_config.jsonRestart Claude Desktop completely for changes to take effect.

Cursor~/.cursor/mcp.jsonRestart Cursor for changes to take effect.

VS Code.vscode/mcp.jsonReload VS Code window for changes to take effect.

Windsurf~/.codeium/windsurf/mcp_config.jsonRestart Windsurf for changes to take effect.

Claude Code.mcp.jsonSave at the project root, then start Claude Code in that project and review the MCP server approval prompt. Keep real credentials out of shared files.

io.github.bolyra/mcp