Summary
✨ AI‑Generated
A forward-deployed engineering role for a Python developer building secure agent runtimes and MCP integrations. You will implement trust checks, identity handling, policy-aware tool behavior, secure gateways, testing infrastructure, and privacy-conscious observability while making complex agent capabilities easy for other engineers to adopt.
Highlights
Highly technical engineering role focused on agent runtimes, trust infrastructure, and Model Context Protocol integrations. Offers substantial ownership over implementation, testing, compatibility, secure identity handling, and developer-facing tooling for emerging agentic systems.
Description
Forward Deployed Engineer – Agent Runtimes & MCP (Python)
This role builds the Agent Trust integration for agentic commerce into Anthropic's Claude Commerce Agents blueprint and the surrounding Claude tooling, with the goal of making it adoptable by merchant engineers in a day.
First 90 days
Stand up the blueprint's verticals on all three runtimes, run its test and verification suites, and pin and maintain fork compatibility.Implement the AgentTrustMixin for the blueprint's StorefrontBackend / MerchantBackend.
This covers identity binding at session start, in-path trust checks on cart and checkout hooks, decision mapping to native tool-result semantics (Unavailable, domain errors), per-hook fail policy, PII-free decision logging, and outcome reporting.Build and test the storefront MCP server path for Managed Agents behind an authenticating gateway, with vault-credential and signed-claim identity options.Write the Agent Trust MCP server (advisory tools) and a companion skill for the blueprint's commerce-builder Claude Code plugin.Author the merchant integration guide and measure merchant effort (files, lines, prompt bytes, gates, time-to-first-verified-session) as an acceptance metric.Must have
5+ years of Python with strong asyncio, typed code, pytest, and ruff, plus comfort subclassing abstract backends and mapping exceptions to structured results.Experience shipping tool-using LLM agents on the Claude Messages API or Agent SDK, and an understanding of prompt caching and why static-prefix bytes matter.Experience building or integrating MCP servers (streamable HTTP) with authentication in front, and an understanding that a model can skip a tool call, so enforcement must live server-side.JWT/JWKS verification, OAuth client credentials, and secrets handling, plus the ability to call a real-time risk API within a latency budget and reason about fail-open vs.
fail-closed.Ability to read unfamiliar, unmaintained open-source code quickly, cite module and line for every claim, and work from written acceptance criteria without supervision.Nice to have
Claude Managed Agents beta (agents, environments, sessions, vaults, always_ask semantics) · Claude Code plugin/skill authoring · Bedrock/Vertex/Foundry routing · E-commerce domain (cart, variants, checkout hand-off) · Node 22 for example storefronts
Deliverables
agent-trust-commerce Python package · Managed Agents gateway variant · Agent Trust MCP server · commerce-builder companion skill · Integration guide · Retail vertical demo
