Install
$ agentstack add mcp-deeplethe-demarche ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
Demarche
[](https://github.com/deeplethe/demarche/actions/workflows/ci.yml) [](https://pypi.org/project/demarche/) [](https://www.npmjs.com/package/@deeplethe/demarche) [](LICENSE)
Vendor-neutral identity verification for AI agents.
Demarche sits between your application and the growing ecosystem of agent-identity issuers — Microsoft Entra Agent ID, Auth0 for AI Agents, WorkOS, OpenAgents, any OAuth On-Behalf-Of issuer. Integrate once, verify agents from any issuer, with no vendor lock-in.
> An agent acts in your démarche: a formal action taken on your behalf, > with verifiable authority. Demarche makes that verification a one-line > integration.
from typing import Annotated
from fastapi import Depends, FastAPI
from demarche import Verifier, VerificationResult, entra_agent_id
from demarche.fastapi import DemarcheAuth
verifier = Verifier(adapters=[
entra_agent_id(tenant_id="", audience="api://your-app"),
])
auth = DemarcheAuth(verifier)
app = FastAPI()
@app.get("/whoami")
async def whoami(
agent: Annotated[VerificationResult, Depends(auth.require_agent)],
):
return {
"user": agent.principal_id, # who delegated authority
"agent": agent.agent_id, # which agent is acting
"scopes": list(agent.scopes),
}
Status
Pre-alpha — v0.0.0 published on PyPI and npm as of 2026-05-18. Public API will likely change before the v0.1.0 cut; pin exact versions if you adopt early. 83 tests passing, ruff clean, mypy strict clean. Architecture contract in [spec/v0.1-architecture.md](spec/v0.1-architecture.md).
Why Demarche
In 2026 the agent-identity ecosystem has fragmented — Microsoft, Auth0, WorkOS, Google A2A, OpenAgents, IETF OAuth OBO drafts, W3C VC — and a scan of ~2,000 public MCP servers found all of them lacked authentication. The protocols exist; integrating N issuers requires N SDKs with N shapes. Demarche pays that cost once on behalf of integrators.
We are to agent identity what Plaid is to bank connections: a neutral normalisation layer over a fragmented ecosystem.
Supported issuers (v0.1)
| Issuer | Module / factory | |---|---| | OAuth OBO (any standards-conformant issuer) | OAuthOBOAdapter | | Microsoft Entra ID / Entra Agent ID | entra_agent_id(tenant_id=..., audience=...) | | Auth0 / Auth0 for AI Agents | auth0_ai_agents(domain=..., audience=...) |
Planned for v0.2+: Google A2A (post-v0.9 signed-Agent-Card freeze), WorkOS, OpenAgents, W3C VC. Track in issues.
Install
pip install demarche # core + OAuth OBO + Entra/Auth0
pip install 'demarche[fastapi]' # + FastAPI integration
npm install @deeplethe/demarche # TypeScript (placeholder; parity in v0.2)
The npm package is scoped as @deeplethe/demarche because npm's typosquat-similarity policy rejected the unscoped demarche name (too close to an existing decache package). The Python package on PyPI is the unscoped demarche.
Quickstart — generic OAuth OBO
from demarche import Verifier, OAuthOBOAdapter, JWKSKeyProvider
verifier = Verifier(adapters=[
OAuthOBOAdapter(
issuer="https://your-issuer.example/",
audience="https://your-api.example",
key_provider=JWKSKeyProvider(
"https://your-issuer.example/.well-known/jwks.json"
),
),
])
result = await verifier.verify(token)
result.principal_id # the user who delegated authority
result.agent_id # the agent acting on the user's behalf
result.scopes # what the agent is authorized to do
result.audit_id # opaque ID for log correlation
FastAPI integration
A complete worked example lives in [examples/fastapi-app/](examples/fastapi-app/) — runnable smoke test included.
from demarche.fastapi import DemarcheAuth
auth = DemarcheAuth(verifier)
@app.post("/book-meeting")
async def book_meeting(
agent: Annotated[
VerificationResult,
Depends(auth.require_scope("calendar.write")),
],
):
return {"booked_by": agent.principal_id, "via_agent": agent.agent_id}
Built-in HTTP error mapping:
| Failure | Status | |---|---| | Missing / malformed Authorization header | 401 | | Token issuer not recognised | 401 | | Bad signature / expired / malformed token | 403 | | Missing required scope | 403 |
Authentication is checked before scope — unauthenticated requests never leak whether a given scope exists.
What Demarche is not
- Not an issuer. Demarche does not mint credentials. It validates
credentials issued by Entra, Auth0, OpenAgents, etc.
- Not an agent framework. Bring your own runtime.
- Not a governance platform. Audit-log persistence and compliance
evidence bundles are out of scope for v0.x (planned for v0.5+).
Roadmap
| Version | Scope | |---|---| | v0.1 | Core verifier + OAuth OBO + Entra/Auth0 profiles + FastAPI integration | | v0.2 | TypeScript SDK at parity; WorkOS profile | | v0.3 | A2A Agent Card adapter (post-spec freeze); Express middleware | | v0.5 | Optional hosted audit-log service; W3C VC adapter | | v1.0 | Third-party security audit |
Threat model
See [spec/v0.1-architecture.md §5](spec/v0.1-architecture.md#5-threat-model-v01). Highlights: confused-deputy mitigation via separated principal_id and agent_id; downgrade-attack mitigation via strict first-match adapter resolution; key-rotation handling via JWKSKeyProvider with refresh rate-limiting.
Packages
| Package | Language | Registry | Status | |---|---|---|---| | demarche | Python ≥3.10 | PyPI | Pre-alpha | | @deeplethe/demarche | TypeScript ≥20 | npm | Placeholder (parity in v0.2) |
License
Apache 2.0. See [LICENSE](LICENSE).
Source & license
This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: deeplethe
- Source: deeplethe/demarche
- License: Apache-2.0
- Homepage: https://demarche.ai
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.