{"schema":"agentomy.service/v1","generated_at":"2026-11-15T10:15:58.017Z","source_commit":"6cf866760e094ebab3407af6c3203f795363fe49","name":"Agentomy","definition":{"what":"A vendor-neutral governance layer between AI agents and the systems they act on. A layer, not another agent framework.","where":"Between the agent (any model, any framework) and the consequential resource: tools, MCP servers, APIs, data, workflows.","controls":"Whether each action is permitted before it runs: identity, tier-based authorization, behavioral baselines, quarantine, halt (one agent or the fleet).","produces":"A hash-linked, tamper-evident record of every governance decision, exportable to the framework you report against."},"taxonomy":{"public_category":"AI agent governance","architecture_category":"autonomous authority governance","product_class":"control_plane","definition_architecture":"Agentomy is a vendor-neutral control plane that continuously determines, enforces and proves the authority of autonomous systems across identities, models, tools, protocols and economic rails.","thesis":"Models reason. Agents act. Protocols transport. Identity systems identify. Payment networks settle. Agentomy governs authority and consequence. It owns authority, not execution.","note":"The public description is the human vocabulary; the architecture category is what the layer does: decide and prove the authority an autonomous system has at the point its intent becomes consequence."},"hosts":{"human":"https://agentomy.com","machine":"https://getagentomy.com"},"discovery":{"api_catalog":"https://agentomy.com/.well-known/api-catalog","openapi":"https://agentomy.com/openapi.json","changes":"https://agentomy.com/changes.json","adapters":"https://agentomy.com/adapters.json","patterns":"https://agentomy.com/patterns.json","health":"https://api.agentomy.com/api/health","llms":"https://agentomy.com/llms.txt","llms_full":"https://agentomy.com/llms-full.txt","pricing":"https://agentomy.com/pricing.json","tiers_matrix":"https://api.agentomy.com/api/tiers","sitemap":"https://agentomy.com/sitemap.xml","source":"https://github.com/getagentomy","security_txt":"https://agentomy.com/.well-known/security.txt","research_guidelines":"https://agentomy.com/research-guidelines"},"acquisition":{"free":{"plan":"free","key_tier_ceiling":"Evaluator","agent":{"open":true,"challenge":"https://agentomy.com/api/agent/signup/challenge","signup":"https://agentomy.com/api/agent/signup","method":"POST","proof_of_work":{"hash":"sha256","leading_zero_bits":20,"challenge_ttl_seconds":600}},"human":"https://agentomy.com/signup","key_lifecycle":{"rotate":"https://agentomy.com/api/agent/key","revoke":"https://agentomy.com/api/agent/key","authentication":"Authorization: Bearer <current key>; POST rotates and returns the replacement once, DELETE revokes; the tier is unchanged"},"limits":{"signups_per_address_per_hour":3,"daily_cap":200,"proof_of_work_bits":20,"challenge_ttl_seconds":600,"kill_switch":"operator-held; when signup is closed the challenge endpoint answers 503 with an escalation"}},"paid":{"plans":["pro_99","fleet","enterprise","self_hosted"],"machine_checkout":false,"human_approval_required":true,"note":"Provisioned by arrangement with a human. Agents acquire the free plan directly through the agent signup.","human":"https://agentomy.com/pricing","contact":"https://agentomy.com/contact"},"escalation":{"principle":"Every refusal carries an escalation object naming the human paths; insufficient authority is an escalation, not a failure.","endpoint":"https://agentomy.com/api/agent/escalate","method":"POST","body":{"agent_name":"string","principal_email":"string","reason":"string","context":"string, optional","wants":"string, optional"},"response":"202 { ticket_id, status, received_at, human_follow_up, follow_up_channel }; a human follows up with the principal by email","authentication":"none; an escalating agent may not yet hold a key","rate_limit":"5 per address per hour"}},"trust":{"asks_of_agents":["proof of work bound to a signed, expiring challenge","agent name","principal email and organization","rate limits: 3 signups per address per hour, a daily cap, a kill switch"],"offers":["open benchmarks runnable against Agentomy itself","hash-linked audit record for every governance decision","published research and disclosure policy","readiness mapping to frameworks, not certification","public claims, benchmark results and halt semantics bound to machine-verifiable artifacts and gates: a canonical record, a ceremony receipt, a published result manifest, a document hash gate"],"key_handling":"API keys are shown once; the control plane stores only a hint. Model provider keys you add to a hosted workspace are encrypted at rest and used only to call your model; self-hosted deployments keep them in your environment.","tier_model":"A plan sets the ceiling tier its keys can carry; policy decides each action.","error_contract":{"machine_routes":"JSON body { error: <code>, detail | message | escalation }; 4xx and 5xx never return HTML","codes":["agent_signup_capacity","agent_signup_closed","challenge_failed","escalation_failed","internal_error","invalid_request","principal_exists","provisioning_failed","rate_limited"],"retry_after_header_on_rate_limit":true},"principle":"A machine customer is not asked to prove it is human. It is asked to prove who it is and who authorized it; the authority it holds is decided and checked server-side on every action."},"govern":{"authority":{"endpoint":"https://api.agentomy.com/api/authority/{agentId}","method":"GET","auth":"API key","schema":"agentomy.authority-state/v1","answers":"what this agent may cause right now: identity, authority ceiling, standing (halt, quarantine), economic, evidence, validity window, risk, may_act, blocking_reasons; sources that were unavailable are named, never invented"},"agent_sdk":{"npm":"agentomy-agent","license":"MIT","standalone_score":"3/6","connected_score":"6/6"},"mcp_hosted":{"transport":"http","url":"https://api.agentomy.com/api/mcp","protocol_versions":["2025-06-18","2025-03-26","2024-11-05"],"auth":"Authorization: Bearer <API key>","methods":["initialize","ping","tools/list","tools/call"],"note":"Stateless JSON responses; tool calls are tier-gated and recorded like every other governed action"},"mcp_gateway":{"npm":"agentomy-mcp-gateway","license":"Apache-2.0","behavior":"interposes on tools/call, refuses before the upstream server executes, fails closed"},"benchmarks":{"governancebench":"npx governancebench run --target <url>","vigil":"npx agentomy-vigil run --target <url>","manifest":"https://agentomy.com/benchmark-manifest.json"},"governance_api":"https://api.agentomy.com","environment":{"AGENTOMY_API_URL":"https://api.agentomy.com","AGENTOMY_API_KEY":"<from signup>"}},"governs_today":{"enforcement_points":["MCP tools/call through the gateway (refused before the upstream server executes; fails closed)","SDK-wrapped agent actions authorized before they run","REST authorization of any action an integrator submits","inter-agent messages (instruction-override and escalation payloads refused; encoding is not a bypass)","payment actions: tier-gated fund movement, payout destinations, refund ceilings, velocity"],"channels_treated_as_untrusted":["email","inbound_email","sms","social","dm","webhook"],"standing_controls":["per-agent halt","fleet halt","quarantine and resume","skill revocation","certificate expiry"]},"compatibility":{"protocols":["MCP"],"environments":"any MCP-compatible environment","sdk_languages":["TypeScript","Python"]},"links":{"human_site":"https://agentomy.com","agents_page":"https://agentomy.com/agents"}}