CoreGuard

AI Governance Platform

EVE is a proof-bearing governance and enforcement platform for AI agents and models. It makes a deterministic policy decision — ALLOW, BLOCK, or MODIFY — before a governed action executes, and it can produce a signed evidence record bound to that decision that you can verify independently, offline, without trusting the EVE server.

The point of difference: the governance verdict is computed by deterministic policy code. No LLM sits in the CoreGuard decision path, so the same input under the same policy produces the same decision.

What a governance platform actually enforces

Pre-execution decisions — a permitted-or-not verdict is returned before a tool call, model action, or MCP request runs, not after the side effect.
Signed decision evidence — each decision can emit an evidence envelope (ECDSA P-384 (AWS KMS) in production configuration, HMAC-SHA256 fallback) so a reviewer can confirm what was decided and that the record was not altered.
Independent verification — evidence verifies in Python, Node, or the browser using the public key, proving authenticity and integrity.
Fail-closed behavior — in embedded, hosted, sidecar, MCP-gateway, and sovereign modes, a service error resolves to BLOCK or raises. No SDK silently falls back to permissive behavior.

SDK example

Python (published on PyPI):

from eve_coreguard import CoreGuardClient

client = CoreGuardClient(api_key="eve_sk_...")  # hosted mode: reaches CoreGuard via base_url

result = client.evaluate(
    request_id="sess_123",
    tenant_id="org_abc",
    proposed_action={"type": "wire_transfer", "amount": 50000, "destination": "acct_9c1"},
    model_output={"decision": "approve"},
    context={
        "principal_id": "agent_007",
        "session_id": "sess_123",
    },
    policy_set="banking_aml_v1",
    include_evidence=True,
)

print(result.decision.status)  # ALLOWED / BLOCKED / MODIFIED
print(result.risk.level)       # risk level: LOW / MEDIUM / HIGH

Verify the returned decision record later, without contacting EVE:

from eve_coreguard import verify_decision_record

report = verify_decision_record(result.signed_governance, public_key_pem=pubkey)
assert report.valid is True   # signature 'kms-ecdsa-p384-valid', schema checked

Readiness

The EVE SDK core is a RELEASE CANDIDATE WITH EXCLUSIONS. The core governance architecture is pilot-validated. Signed evidence and offline verification are SUPPORTED. The Python client (eve-coreguard 0.2.7) and TypeScript client (eve-ai-governance 0.3.1) are release-candidate core clients. The clients are published on public registries: pip install eve-coreguard (PyPI) and npm install eve-ai-governance (npm).

Limitations

The pip client governs via hosted mode and needs a reachable endpoint; embedded in-process governance is the EVE service itself, not the client wheel.
Offline verification proves that evidence is authentic and unaltered — it does not attest that the underlying decision was correct.
The HMAC-SHA256 fallback signature is symmetric and is not independently verifiable; independent verification requires the ECDSA P-384 public key.
"No LLM in the decision path" applies to the governance verdict only; the governed application may still use LLMs.

Next step

Request a pilot to run EVE against your own policies and agents. You keep the evidence, and you verify it yourself.