Superposition now works inside your AI tools

By Xiang LiMay 20, 2026
Superposition now works inside your AI tools

Today we're launching MCP support in Superposition. It's bidirectional, and we believe it's the first recruiting platform to ship this.

We built Superposition to be the best AI headhunter on the planet, but even the best recruiter is limited if they can only work inside their own office. Superposition now plugs directly into the AI tools you're already using to build your company, and your favorite tools plug right back into Superposition.

Connect Superposition to your AI tools

Superposition as an MCP server means external AI tools can now connect to your workspace. Ask Claude Code, Cursor, or Codex to pull pipeline metrics, search candidates, summarize an interview, or draft outreach. Your recruiting data lives inside whatever agent you're already using.

Superposition as an MCP client means our agent calls the tools you already have. Connect Notion, Linear, or a custom endpoint. When a new role opens, the agent can pull the hiring brief automatically.

When you connect an external client, it gets read access to:

  • Candidate search, profiles, and contact history (full email and Slack threads)
  • Jobs, pipeline metrics, and interview summaries
  • Client health, agreements, and billing status

Writes route through Superposition's Agent. Stage updates and outreach drafts stay inside our review layer before anything goes out.

How it works

Settings → Connect MCP → pick the feature groups you want to expose → copy your preset URL → paste it into Claude Code, Cursor, or Codex. OAuth handles consent.

Each preset is scoped and revocable from Authorized Apps.

Use cases you can start with

1. You ask “How's our Founding Engineer pipeline looking?” and get a full snapshot with stage counts, recent activity, and a flag on your longest stalled candidate.

Claude Code pulling a Founding Engineer pipeline snapshot from Superposition

2. You follow up on that candidate and get their complete profile, GitHub activity, application history, and a recommended next action.

Candidate profile pulled into Claude Code with stage, source, and activity

3. You ask “Who's been stuck waiting to schedule an interview for more than a week?” and get names, wait times, and an offer to draft outreach to both.

Two candidates flagged as blocked on scheduling, with wait times and outreach offer

4. You ask “What themes are showing up in candidate rejections this month?” and get the top three reasons with percentages, plus a suggestion on how to fix the biggest one.

Top three rejection themes for Founding Engineer role with calibration suggestion

All of this happens inside Claude Code, Cursor, or Codex. You never have to open Superposition.

You control exactly what gets accessed

Every connection uses a preset that you configure. You decide which features a client can use.

Maybe your Claude Code setup can search candidates and view jobs but can't send outreach emails. Maybe your team's shared setup has full access. You can create separate presets for different people or tools, and revoke any of them instantly. Tokens are encrypted, every request needs a valid short-lived token, and a copied URL alone won't get anyone in. Your candidate data doesn't become a free-for-all just because you connected a new tool.

How this improves your workflow

Most recruiting tools are closed loops. You run searches inside their UI, copy results out, and paste them into whatever you're actually using. MCP breaks that. Superposition becomes a tool your agents call, not a tab you switch to.

Get started

Existing users: Settings → Connect MCP.

New to Superposition? Book a demo and we'll walk you through it.

See Superposition inside your AI stack.
Hook our recruiter into Claude Code, Cursor, or Codex in minutes.

Written by

Xiang Li
Xiang LiCTO & Co-Founder, Superposition

Building startups since college. On the founding team of Vitable Health (YC S20), helped take the company from pre-seed to Series A, past $10M ARR, and hundreds of thousands of patients. Writes about AI agent architecture and automating knowledge work.