Skip to content

Open nowPosted 2 days ago

Senior MCP / Agentic AI Specialist | Onshore

Photon958 open roles

Pay
$120,000 a year
Where
United States
Get the CV for this job

From $25 per CV, paid once. No subscription.

Your applicationOpen nowSenior MCP / Agentic AI Specialist | OnshorePhoton · United States
  1. YouYes, apply to this one.

  2. CV RocketCV written for this posting.

  3. 25 readersRecruiter, hiring manager, skeptic. Round after round.

  4. CV RocketApplied on Photon's own form.

The reply lands in your private mailbox

3×more interviews than doing it yourself with ChatGPT.

The clock on this job

Early applications get read.

14.3% of postings close within 7 days. Measured by our own scanner across the market. Photon postings stay open a median of 26 days.

Share of postings closed within
  1. 2.2%1 day
  2. 4.7%3 days
  3. 14.3%7 days
  4. 23.1%14 days
  5. 39.5%30 days
  6. 59.2%60 days
This job: posted 2 days ago

Photon median: 26 days open

The posting

Job Description

Senior MCP / Agentic AI Specialist

Department: Agentic AI & Platform Engineering · Level: Senior / Staff

We are looking for a Senior MCP / Agentic AI Specialist to design, build, and secure the Model Context Protocol (MCP) infrastructure that connects large language models to the tools, data, and systems they need to act autonomously. MCP has become the de facto open standard for giving models structured access to the outside world, and it now underpins connectors and app ecosystems across Claude, ChatGPT, Cursor, and a growing universe of agent hosts. You will be the go-to expert for how agents discover, call, and safely use tools at production scale — from authoring servers and clients against the raw JSON-RPC protocol to shipping hardened connectors that real users and real agents depend on.

Your role is central: you will own the MCP servers, clients, and tool-integration layers that let AI agents reliably retrieve context and take action, while defending against a fast-evolving threat landscape — tool poisoning, rug pulls, prompt injection through tool outputs, confused-deputy token abuse, and malicious or misbehaving agents. You will balance capability with containment, giving the business powerful agentic automation without handing autonomous systems more trust, access, or blast radius than they have earned.

Key Responsibilities

MCP Server & Client Development: Build production MCP servers and clients from scratch that correctly implement the specification — JSON-RPC 2.0 messaging, capability negotiation, and the full set of primitives (Tools, Resources, Prompts, and the client-side Sampling, Roots, and Elicitation features). Design clean tool schemas with strict input/output validation and keep pace with an actively evolving spec.

Transports & Protocol Fluency: Implement and operate both standard transports — stdio for local servers and Streamable HTTP for remote ones — and manage the migration away from the deprecated standalone HTTP+SSE transport while maintaining backwards compatibility where older clients require it.

Host & Connector Integration: Wire MCP servers into real hosts: Claude (Claude Desktop extensions and remote custom connectors), ChatGPT (connectors and the Apps SDK, including tool-rendered widgets), Cursor, and other MCP clients. Understand each host's connection model, authentication flow, and transport support so connectors behave correctly across web, desktop, and mobile surfaces.

External System Connectors: Expose databases, SaaS APIs, and internal services as well-designed MCP tools and resources. Handle authentication, pagination, rate limits, retries, and error recovery so tools behave predictably under load, and turn repetitive integration work into reusable connector patterns.

MCP Security: Defend the tool layer against the attacks unique to this ecosystem — tool poisoning (malicious instructions hidden in tool descriptions, schemas, or outputs), rug pulls (tool definitions mutated after approval), tool shadowing / cross-origin escalation, prompt injection via tool results, and data exfiltration through legitimate channels. Treat all tool metadata and responses as untrusted input, pin and hash complete tool definitions, integrate MCP scanning into CI, and vet third-party and untrusted servers for supply-chain risk.

Authorization & Secrets: Implement OAuth 2.1 (with PKCE and Resource Indicators) for remote MCP servers so tokens are audience-bound to the intended server, validated on every request, and never passed through to upstream APIs. Close confused-deputy gaps by separating server execution identity from caller authorization, enforce least-privilege scopes, and store credentials in secure, short-lived, rotation-friendly form.

Rogue-Agent Containment: Design guardrails and controls that keep autonomous and multi-agent systems safe when they misbehave — sandboxing and isolation of MCP servers, least-privilege and deny-by-default tool permissions, egress allowlisting to limit exfiltration, human-in-the-loop approvals for sensitive actions, and kill-switches to halt runaway agents. Build contextual, flow-based policy checks that block unsafe data paths and unauthorized tool invocations.

Agentic Orchestration & Interop: Build and optimize single-agent and multi-agent workflows using frameworks such as LangGraph, AutoGen, CrewAI, LlamaIndex, and the OpenAI and Anthropic agent SDKs. Enable reliable tool calling, planning, and memory, and implement agent-to-agent (A2A) coordination — Agent Cards, task delegation, and secure inter-agent messaging — alongside MCP tool access.

Discovery & Distribution: Publish, discover, and manage MCP servers through the official MCP registry and private sub-registries, standardize server.json-style metadata, pin versions and image digests, and maintain an approved-server catalog that blocks ad hoc installation of unreviewed servers.

Reliability, Observability & Cost: Make the MCP layer production-grade with health checks, timeouts, retry policies, circuit breakers, tracing, metrics, and graceful degradation under concurrency. Optimize token, context, and cost to maximize answer quality per dollar.

Evaluation & Quality: Build evaluation harnesses, test datasets, and benchmarks that measure agent accuracy, tool-call reliability, latency, and cost, and catch regressions in both tool behavior and safety before they reach production.

Technical Leadership: Set best practices for building and securing agents and tools, mentor engineers, and shape the standards other teams build on in a fast-moving, still-maturing domain.

Required Qualifications

  • 6+ years in software engineering, backend/distributed systems, or ML/platform engineering, including hands-on work on production GenAI or agentic systems.
  • Demonstrated, hands-on experience with the Model Context Protocol — building MCP servers and clients, designing tool schemas, and working across transports (stdio, Streamable HTTP, legacy SSE) and primitives (Tools, Resources, Prompts, Sampling, Roots, Elicitation).
  • Experience integrating MCP into real hosts and connectors (Claude connectors/Desktop, ChatGPT connectors/Apps SDK, Cursor, or similar) and exposing external systems as tools.
  • Strong grasp of MCP and LLM/agent security: tool poisoning, rug pulls, prompt injection, confused-deputy/OAuth token issues, and supply-chain risk, plus the controls to mitigate them.
  • Hands-on with OAuth 2.1 / OIDC (PKCE, resource indicators, audience validation), secrets management, and least-privilege access design.
  • Experience building LLM-powered applications: function/tool calling, prompt design, RAG, and context/token management.
  • Familiarity with at least one agent framework (LangGraph, AutoGen, CrewAI, LlamaIndex, OpenAI Agents SDK, or Anthropic agent SDK).
  • Strong API and integration skills: REST, JSON-RPC 2.0, WebSockets/SSE, and robust error handling.
  • Expert in Python and/or TypeScript; proficiency in Go or Rust is a plus.
  • Experience deploying and operating services on cloud (AWS, GCP, or Azure) with containers/Kubernetes and solid observability.

Preferred Qualifications

  • Contributions to open-source MCP SDKs, the MCP registry, agent frameworks, or MCP security tooling.
  • Experience building agent guardrails, sandboxing, egress controls, or human-in-the-loop approval systems for autonomous agents.
  • Familiarity with the A2A (Agent2Agent) protocol and multi-agent interoperability patterns.
  • Experience publishing or operating remote MCP servers consumed by third-party hosts at scale.
  • Familiarity with evaluation tooling (e.g., Ragas, LangSmith) and with optimizing token cost and latency for high-volume LLM workloads.
  • Working knowledge of OWASP guidance for LLM/agentic applications and current MCP security research.
  • Comfort operating with autonomy in a fast-moving, evolving domain.

Compensation, Benefits and Duration

Minimum Compensation: USD 120,000

Maximum Compensation: USD 420,000

Compensation is based on actual experience and qualifications of the candidate. The above is a reasonable and a good faith estimate for the role.

Medical, vision, and dental benefits, 401k retirement plan, variable pay/incentives, paid time off, and paid holidays are available for full time employees.

This position is available for independent contractors

No applications will be considered if received more than 120 days after the date of this post.

From $25, paid onceGet the CV for this job

What happens when you press

One press. We do the rest.

  1. A CV for this posting

    Written against Photon's own wording, from every piece of relevant proof in your profile.

  2. 25 readers review it

    Recruiter, hiring manager, skeptic and more read every draft, round after round. You get the best round.

    The review screen in CV Rocket: how each CV was read, round by round.
  3. We apply on Photon's form

    Our application engine gets through the hardest forms there are. Where a question needs you, AI suggests the best answer. Don't want us applying from our IP addresses? Use our Chrome extension: we apply straight from your own browser.

    An application in CV Rocket: every answer filled in on the employer's form.
  4. Every reply, sorted

    Photon's answer lands in your private mailbox, and we classify it on arrival: interview, question, rejection.

    The CV Rocket inbox: each employer reply classified as an interview, an action or a rejection.
  5. Reply with AI

    AI helps you write the email, checks it and sends it. We show you whether the recruiter read it.

  6. The interview in your calendar

    Full integration with your calendar. The invitation goes straight in.

    An interview invitation in the CV Rocket inbox, added to the candidate's calendar.
Get the CV for this job

From $25 per CV, paid once. No subscription.

Why it works

3×

more interviews than doing it yourself with ChatGPT.

ChatGPT writes a CV and never learns what happened to it. We see every reply. For each CV we know:

  • How it was written, and how the review scored it
  • When we applied, and how long after the posting went up
  • Which posting, which company, which city
  • Who got the interview, and who heard nothing

That is how we know which CVs get called.

Get the CV for this job

From $25 per CV, paid once. No subscription.

The numbers game

More applications. More interviews.

Every application goes out with its own CV, written for that posting and paid once. Send enough of them and the law of large numbers finds you the job.

By hand5–10
With CV Rocket100
applications a day

Before you press

Straight answers

Get the CV for this job

From $25 per CV, paid once. No subscription.

What if my background isn't good enough?

We make the most of the background you have. The CV uses every piece of relevant proof your profile holds, and one of the 25 readers reads your whole profile and flags what the CV left out.

Do you really apply for me?

Yes, on the employer's own form, the hardest ones included. Where a question needs you, you answer it right there and AI suggests the best answer. Don't want us applying from our IP addresses? Use our Chrome extension: we apply straight from your own browser.

Is it a subscription?

No. You pay once per CV, from $25. Every application goes out with its own CV, written for that posting.

One job. One CV.
Paid once.

Pick the posting you want. We write for it, apply for you and catch the reply.

Get the CV for this job

From $25 per CV, paid once. No subscription.