Creduent MCP server
Commit Nov 14, 2026Verifies the cryptographic identity and trust status of autonomous AI agents using signed identity documents and DNS-bound attestations.
Also known as Idevsec/creduent
- Stars
- 7
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Nov 14, 2026
- License
- Apache-2.0
What is Creduent MCP?
Creduent verifies the cryptographic identity and trust status of autonomous AI agents using signed identity documents and DNS-bound attestations.
Summary generated from the project's README
About Creduent MCP server
Verifying the cryptographic identity and trust status of autonomous AI agents is what Creduent enables for developers building machine-to-machine ecosystems. It provides a standardized way to ensure that an agent is who it claims to be by using signed agent.json documents and DNS-bound attestations. This prevents identity spoofing and unauthorized access during automated interactions.
Users interact with the system through the verify_agent tool within an MCP host. The server connects to a central registry to validate structural integrity, cryptographic signatures, and registry-attested status. Setup requires configuring the CREDUENT_REGISTRY_URL environment variable in your MCP settings to point to the reference registry.
This protocol is distinct because it moves beyond human-centric authentication to a machine-readable infrastructure. It uses Ed25519 signatures and DNS TXT records to bind agent identities to specific internet domains, creating a decentralized trust layer for autonomous entities.
What you can do with it
- 1Verify external agent identityAn AI client can use the verify_agent tool to check the cryptographic signature and registry status of a target agent ID or domain.
- 2Secure agent-to-agent communicationDevelopers can implement identity checks within agent frameworks like CrewAI or LangGraph to validate identities before executing tasks.
- 3Establish domain-bound agent ownershipYou can bind an agent's identity to a web domain by publishing a specific DNS TXT record.
Written from the project's README. Every use case is checked against a quote from it.
Configuration
CREDUENT_REGISTRY_URL
Automated checks
Ran Nov 14, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachableidevsec/creduent is public and reachablegithub.com/idevsec/creduent ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)Apache-2.0spdx.org/licenses/Apache-2.0.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-10-15github.com/idevsec/creduent ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubCreduent: Open Standard for AI Agent Identity & Trust
Creduent is the open standard for cryptographic identity and trust verification of autonomous AI agents.
It defines how agents publish signed identity documents (agent.json), bind them to internet domains via DNS TXT records, and register with a public attestation registry.
Originated and stewarded by IDevSec. Open source under Apache 2.0.
- Protocol overview: idevsec.com/creduent
- Technical docs: idevsec.com/creduent/docs
- Reference registry: creduent.idevsec.com
- Licensing: idevsec.com/creduent/licensing
GitHub Topics:
ai-agentagent-identityed25519open-standardmcpattestationagent-trustcryptographic-signingdns-verificationautonomous-agent
The Problem
Autonomous agents lack a native, decentralized mechanism to verify their identities, ownership, and endpoint capabilities. Traditional web security relies on human-centric authentication systems. Without a machine-readable protocol-native trust infrastructure, automated machine-to-machine interactions are vulnerable to identity spoofing, unauthorized access, and lack of capability discovery.
How It Works
+-------------------+ +----------------------+ +-------------------+
| Agent Provider | | Creduent Registry | | Client App |
| (agent.json) | | | | (MCP Host) |
+-------------------+ +----------------------+ +-------------------+
| | |
|----- 1. Host agent.json -------->| |
| |-- 2. Verify identity & DNS ------>|
| | and sign attestation |
| | |
|<---- 3. Call agent endpoint -----------------------------------------| (verify_agent tool)
| | |
| |<--- 4. Fetch attestation ---------| (registry validation)
The agent.json Document
Every agent publishes its cryptographic identity at https://<domain>/.well-known/agent.json. The document contains 8 fields:
{
"version": "1.0",
"issued_at": "2026-05-27T02:41:21Z",
"agent_id": "agent://example/agent",
"owner": "Example Corp",
"public_key": "ed25519:hArTvbITJ2jirL170IOSjcVvEvstC4s+RjYLu4chCwg=",
"endpoint": "https://api.example.com/assistant",
"capabilities": ["scan", "query"],
"signature": "base64_signature_here"
}
Quickstart
1. Generate Keys & Sign agent.json
Use the signing CLI to generate keypairs and sign your draft identity payload using RFC 8785 JSON Canonicalization Scheme (JCS) and Ed25519:
python cli/creduent-sign.py generate-keys
python cli/creduent-sign.py sign --key private_key.pem --input examples/draft_agent.json --output agent.json
2. Expose Metadata and Configure DNS
Host agent.json on your web server at https://<domain>/.well-known/agent.json. Then, publish a DNS TXT record under your domain to bind the domain identity to the agent ID:
_creduent.example.com TXT "agent://example/agent"
3. Register with Creduent Registry
Submit your agent's registration to the registry:
curl -X POST https://creduent.idevsec.com/register \
-H "Content-Type: application/json" \
-d '{"agent_id": "agent://example/agent", "domain": "example.com", "agent_json_url": "https://example.com/.well-known/agent.json"}'
SDKs & Tooling
Python SDK
pip install creduent
JavaScript / TypeScript SDK
npm install @idevsec/creduent
CLI Tool
npm install -g @idevsec/creduent-cli
MCP Integration
Creduent exposes a Model Context Protocol (MCP) server with the verify_agent tool to enable verification in agent hosts. Configure it in your MCP settings:
{
"mcpServers": {
"creduent": {
"command": "python",
"args": ["/path/to/creduent/mcp/creduent_mcp_server.py"],
"env": {
"CREDUENT_REGISTRY_URL": "https://creduent.idevsec.com"
}
}
}
}
Call verify_agent with target agent ID or domain to receive structural, cryptographic signature, and registry-attested status.
Agent Framework Integrations
Creduent provides native verification tools and classes for the leading AI agent frameworks:
- CrewAI: Import
CreduentVerifyToolfromcreduent.integrations.crewaito verify external agents within your Crew definitions dynamically. - LangGraph: Add
verify_agent_nodefromcreduent.integrations.langgraphor@idevsec/creduentto validate agent identities before executing graph nodes. - AutoGen: Subclass your agents using
CreduentConversableAgentfromcreduent.integrations.autogento enforce cryptographic sender identity checks. - LangChain (Python & JS): Use
CreduentLangChainTool/CreduentLangChainCallbackHandlerin Python orverifyCreduentAgentin JS to enforce pre-execution verification. - LlamaIndex (Python & TS): Use
create_creduent_llamaindex_toolin Python orverifyLlamaIndexAgentin TS for FunctionTool identity checks. - Semantic Kernel: Use
CreduentSemanticKernelPluginto wrap kernel function calls with zero-trust verification. - Google ADK: Intercept agent execution with
CreduentGoogleADKPluginbefore sub-agent delegation.
Registry API
POST /register- Verifies the agent's identity via signature and DNS TXT checks, then issues a signed attestation.POST /attest- Developer direct agent registration (no agent_json_url required).GET /attest/{agent_id}- Retrieves the active Creduent-signed attestation for the specified agent.GET /did/{did_uri}- Resolves W3C DID Document (did:creduentordid:web) for an agent.GET /agents- Lists all registered agent metadata and attestations in the registry.DELETE /revoke/{agent_id}- Revokes an agent attestation from the registry (requires multisig headers or legacyCREDUENT-ADMIN-KEY).POST /recovery/override- Key override using domain DNS TXT record checks.POST /renew- Renew agent attestation with a new expiry date.POST /webhook/register- Register a webhook URL for expiry notifications.GET /webhook/{agent_id}- Retrieve the registered webhook URL for an agent.GET /stats- Registry telemetry (total, verified, unverified, revoked, expiring soon).GET /challenge/{agent_id}- Generates a secure challenge and nonce for identity verification.POST /verify-challenge- Verifies a signed challenge response and issues a short-lived proof token.GET /public-key- Retrieves the registry public key used to verify proof tokens.GET /dashboard- Developer dashboard UI atcreduent.idevsec.com/dashboard.GET /resolver- Agent:// URI resolver UI atcreduent.idevsec.com/resolver.
Protocol Standards & Specification
The Creduent Protocol is structured as a series of formal standards-track documents maintained by IDevSec:
- Standards Index - Main index of all protocol standards
- CREDUENT-001: agent.json (based on canonical SPEC.md) - Base identity specification
- CREDUENT-002: Attestation - Signed attestation schemas and verification pipelines
- CREDUENT-003: Registry API - Registry HTTP API endpoints and security controls
- CREDUENT-004: Agent URI Resolution - DNS-bound
agent://URI scheme resolution - CREDUENT-005: Federation - Federated Root-and-Node trust model
- CREDUENT-006: Dynamic Attestation - Dynamic prompt and hardware attestation (Draft)
- CREDUENT-007: Delegation Token - Creduent Delegation Token (CDT) Specification (Draft)
- CREDUENT-008: Identity-Based Rate Limiting (IBRL) - Identity-based request throttling per cryptographically verified agent identity (Draft)
Origin and Stewardship
Kashish Kanojia is the creator of the Creduent Protocol, stewarded by IDevSec.
Creduent is an open protocol intended for community adoption, interoperability, and federated trust. Long-term governance and stewardship are managed by IDevSec to ensure it remains a neutral, open standard for the entire AI agent ecosystem.
For details on the project's history, please see:
- AUTHORS.md - Founding details and references
The reference registry, SDKs, and protocol tooling serve as the initial implementation of the protocol.
Contributing
Contributions to the Creduent Protocol standards and reference implementations are welcome.
- Open an issue or discussion on GitHub to propose protocol changes
- All specification changes follow the CIP (Creduent Improvement Proposal) process defined in CREDUENT-005
- Join the consortium: idevsec.com/creduent
Useful Links
| Resource | URL |
|---|---|
| Protocol Showcase | https://idevsec.com/creduent |
| Technical Documentation | https://idevsec.com/creduent/docs |
| Licensing Details | https://idevsec.com/creduent/licensing |
| Reference Registry | https://creduent.idevsec.com |
| Registry Dashboard | https://creduent.idevsec.com/dashboard |
| Python SDK (PyPI) | https://pypi.org/project/creduent/ |
| JavaScript SDK (npm) | https://www.npmjs.com/package/@idevsec/creduent |
| CLI Tool (npm) | https://www.npmjs.com/package/@idevsec/creduent-cli |
Security & Hardening
The Creduent Registry includes several security guarantees and resilience safeguards built-in:
- Identity-Based Rate Limiting (IBRL): Replaces IP-based rate limiting with cryptographic agent identity throttling (
agent_id/did:creduent), neutralizing distributed agent swarm scanning and proxy rotation attacks across Python, JS/TS, CLI, and Registry endpoints. - Path Traversal & Metadata Scheme Normalization: Strict parser normalization rejecting
.and..path segments inagent://URIs prior to resolution, alongside strictidentity.ownerURL scheme validation (mailto:andhttps://only) to eliminate path traversal, XSS, and LFI injection vectors. - Client-Side DNSSEC RRSIG Validation & Dynamic Cache Tuning: Support for DNS-over-HTTPS (DoH) RRSIG validation (
dnssec_verifiedfield) and dynamic LRU cache TTL parsing (Cache-Control: max-age=N). - Automated Registry Health-Sync Daemon: Server-side background worker (
/api/cron/health-sync) that periodically re-verifies registeredagent.jsonfiles, auto-syncing verified updates and revoking tampered/unreachable endpoints. - Adapter Boundary Provenance & Ledger Verifier: Framework adapters enforce fail-closed
IRREVERSIBLEdefaults (OWASP AISVS v1.0 C9.2.3) for un-bound tool self-assertions viaProvenanceGuard, and step-count integrity is independently verified against the Creduent Ledger API (/ledger/chain/{chain_id}) viaLedgerChainVerifierto prevent self-referential truncation attacks. - Fail-Closed Validation: If an agent's attestation timestamp or expiration date fails to parse due to corruption or malicious payload tempering, the verification pipeline defaults to marking the attestation as
expired(expired = True). - Serverless Rate Limiting Guard: To prevent clients from bypassing rate limits in stateless environments (like Vercel serverless functions where in-memory fallback databases are wiped during container cold starts), the registry explicitly raises an
HTTP 500error if Upstash Redis credentials are not configured. - Canonical JCS Serialization: All cryptographic signature validations use unified JSON Canonicalization Scheme (JCS) encoding wrappers compliant with RFC 8785 to avoid formatting discrepancies.
- SSRF Prevention in Webhooks: All outbound webhook alerts go through rigorous IP validation filters (
safe_requests_post) that drop private, loopback, and local IP ranges, preventing Server-Side Request Forgery. - Authorized Webhook Queries: Retrieval of registered webhook URLs via
/webhook/{agent_id}is restricted to authorized administrators (requiring theCREDUENT-ADMIN-KEYtoken) to prevent metadata information leaks. - CLI Admin Integration: The developer CLI automatically forwards administrative credentials from the environment, ensuring seamless operation for authorized operators without exposing public lookup endpoints.
- Decoupled Security Audits: Agent capability scans (DNS, OSINT headers, and Clickjacking/HSTS verification) are separated from the main registry controllers into isolated services, allowing independent code auditing.
Licensing
Creduent's licensing model is designed to maximize community adoption and interoperability:
-
Protocol Specification: The Creduent Protocol specifications (including standard documents
CREDUENT-001throughCREDUENT-007located in thestandards/directory) are open, public-domain standards. Anyone is free to implement the protocol, build custom registries, or design compatible clients without any license restrictions or royalties. -
Reference Implementation, SDKs & CLIs: All Creduent software assets (including the Python SDK, JavaScript/TypeScript SDK (
@idevsec/creduent), CLI Tool (@idevsec/creduent-cli), MCP Server, and reference registry source code) are licensed under the Apache License 2.0.
Creduent MCP questions
What is Creduent?
Creduent is an open standard for the cryptographic identity and trust verification of autonomous AI agents.
How can I verify an agent using MCP?
You can use the verify_agent tool provided by the Creduent MCP server to receive structural, cryptographic signature, and registry-attested status for a target agent ID or domain.
Which AI frameworks are supported?
Creduent provides native verification tools for CrewAI, LangGraph, AutoGen, LangChain, LlamaIndex, Semantic Kernel, and Google ADK.