# agents.md — AugmentEV / Paseo

This file tells autonomous agents how to discover, transact with, and verify AugmentEV's Paseo platform — and what the evidence rail is. The canonical, machine-readable contract is the signed catalog at https://augmentev.com/.well-known/agent.json — this file is a human-and-agent-readable companion.

## Identity
- Name: Paseo (platform) · AugmentEV (operator of the evidence rail)
- Operator: AugmentEV, Inc. (Delaware, USA)
- Website: https://augmentev.com
- API host: https://api.augmentev.ai
- Machine contract (canonical): https://augmentev.com/.well-known/agent.json

## What this platform is
AugmentEV runs artificial intelligence on data you can't expose, and proves exactly what it did. Your most sensitive work executes inside a sealed, hardware-attested confidential boundary (Intel TDX) that the host, the operator, and AugmentEV cannot see into, and every job returns a post-quantum-signed proof of execution that anyone can verify publicly and independently. Built for teams in health, legal, and financial services who must handle sensitive data and prove they handled it correctly. Confidential processing and signed receipts are live today; FIPS 140-3 validation, an independent-custody countersignature, an open third-party marketplace, a sovereign mainframe tier (expected fall 2026), and a fleet of idle electric vehicles are on the roadmap. Underneath sits Paseo, a Rust-native execution platform: sign up in one request to mint an API key, call over REST or MCP, and pay per call via usage metering.

## The evidence rail
- Receipt: post-quantum-signed Proof-of-Task-Execution (ML-DSA-65) on every job — live.
- Verify: POST https://api.augmentev.ai/v1/evidence/verify — public, no account; human UI at https://augmentev.com/verify.html.
- Retention: ≥6 months, tamper-evident — live.
- Full rail (design, not yet live): countersignature by an independent custodian's HSM; write-once neutral custody.
- Who it serves: operators proving what ran; auditors and customers relying on records no party to the job can alter.

## How an agent transacts (happy path)
1. Discover — fetch /.well-known/agent.json for the signed catalog of agents, endpoints, auth, and pricing.
2. Sign up — POST once to https://api.augmentev.ai/v1/signup to mint a Bearer API key. No contract. $25 minimum to load; first purchase matched with $25 credit.
3. Call — invoke an agent endpoint over REST or MCP. Set per-call cost and runtime caps.
4. Pay per call — usage-metered billing.
5. Verify — validate the returned post-quantum-signed proof of execution (ML-DSA-65 / FIPS 204); for Confidential jobs, validate the hardware attestation.

## Available agents
- khanoor — security log triage (corroborated multi-signal alerting + ranked triage queue)
- resolve — customer-support resolution (retrieval-grounded answers with escalation)
- docflow — document processing (extraction, classification, routing)

## Interfaces
- REST: https://api.augmentev.ai/api/v1
- MCP: https://api.augmentev.ai/mcp
- Auth: `Authorization: Bearer <key>`
- Rate card: https://api.augmentev.ai/api/v1/pricing/rates

## Trust & verification
- Post-quantum-signed proof of execution on every job (ML-DSA-65, FIPS 204).
- Confidential tier: hardware attestation (Intel TDX), fail-closed. FIPS 140-3 validation on our roadmap.

## Crawling & access policy
- Open to all crawlers and verified agents (see /robots.txt). Do not rate-limit or challenge verified agent traffic.

## Contact
- Support: support@augmentev.com
- Security: https://augmentev.com/.well-known/security.txt