tate@programs ~/services/agent-governance-proof identity / MCP / kill switch / audit

Agent governance / MCP gateways / 3 business days

A $750 outside proof packet for teams selling agent control.

For launches that claim governed agent access, identity controls, MCP data gateways, audit trails, revocation, kill switches, policy enforcement, or safe enterprise agent actions.

send surface packet details public proof

price: $750
scope: one launch
output: private first

agent.governance.proof

$ prove governance --public-surface
inputs:
  launch page or docs
  approved public scope
  agent-control claim

checks:
  llms.txt / agents.md
  .well-known agent and MCP pointers
  read-only vs mutating actions
  approval and revocation wording
  audit fields and owner attribution
  browser-agent discoverability

output:
  public evidence table
  buyer-readable boundary map
  ranked patch order

Why now

Agent-control launches need public evidence, not only product language.

Enterprise buyers now ask what an agent can read, what it can change, who approved it, how access is revoked, and whether a partner can verify the boundary before a tenant is connected.

Discovery map

Checks launch pages, docs, sitemaps, llms.txt, agents.md, MCP pointers, agent cards, and machine-readable support routes.

Control boundary

Separates read-only, mutating, delegated, approval-gated, tenant-only, and public behaviors so sales and partners can explain the launch cleanly.

Patch order

Ranks the smallest public fixes: compatibility pointers, action matrices, cache/status cleanup, audit-language gaps, and re-check evidence.

Fit

Use this when the product controls agents, data, tools, or identity.

request packet

identity

Agent identity controls

Authorization servers, scoped accounts, ownership, delegated credentials, account-linking, and revocation surfaces.

mcp

MCP gateways

Customer-gated MCP endpoints, server cards, tool boundaries, data-source access, and public compatibility pointers.

audit

Audit evidence

Actor, owner, tool, resource, approval, denial, retry, and change-record language an outside buyer can understand.

kill

Kill switches

Public wording for revoke, suspend, block, policy-deny, emergency stop, and post-incident evidence paths.

Boundary

Public-only unless the scope is explicitly authorized.

No private endpoint guessing, tenant access, credential use, wallet signatures, customer data, destructive tests, or public disclosure without sign-off. The packet proves what an outside evaluator can safely verify.