A team reviewing performance check-ins around a table

Claude + Lattice: What the Integration Can (and Can't) Do in 2026

Lattice actually shipped an official MCP server — so yes, Claude can read your goals, reviews, feedback, and 1:1s. Lattice launched Lattice MCP on July 7, 2026, billing itself as the first performance-management platform to put an official Model Context Protocol connector in the frontier AI marketplaces. It works with Claude, ChatGPT, and Slack, and it respects the same login and permissions each user already has in Lattice. If you don’t have (or can’t provision) that connector, there’s also a community MCP server and Lattice’s public REST API. But all of these share one hard limit: nothing fires on a Lattice event — no review going overdue, no goal slipping, no 1:1 getting skipped — and there’s no inbox for work to land in.

Here’s each route, when to pick it, and what to use when you want Lattice work that happens without you at the keyboard.


Route 1: Lattice’s official MCP connector (the real answer)

This is the one to reach for first if you can. Lattice hosts a remote MCP server at https://api.latticehq.com/mcp, and it’s added to Claude as a custom (remote) connector over OAuth — no self-hosting, no code.

Setup is a two-part, admin-led job:

  1. In Lattice, a workspace admin goes to Admin → Platform → MCP Server, clicks Add MCP Server, copies the Client ID, and pastes in the redirect URLs Claude provides.
  2. In Claude, an org admin opens Organization Settings → Connectors, chooses Add → Custom → Web, enters the URL https://api.latticehq.com/mcp, opens Advanced Settings, pastes the Client ID (leave Client Secret blank), and saves.
  3. Each user then connects their own Lattice account through the connector, entering their subdomain and logging in — so Claude only ever sees what that user is allowed to see in the Lattice UI.

Once connected, managers can draft, edit, and even submit performance reviews inside the conversation, pulling in goals, feedback, 1:1 notes, updates, and prior reviews — and combining that with context from other connected tools. Two honest caveats: it’s an admin-provisioned, org-level connector (not something an individual on a free Claude plan just flips on), and at launch it’s available to US data-residency customers, with EU support slated for fall 2026.


Route 2: A community MCP server (for individuals or self-hosting)

If you’re not an admin — or you want to run the server yourself against the API — there’s a community option. EoinFalconer/lattice-mcp is a small open-source MCP server that wraps the Lattice Talent API and exposes about 30 tools across users, goals, goal updates, departments, feedback, review cycles, reviewees, and questions. It’s new and low-profile (near-zero stars as of mid-2026), so treat it as a build-it-yourself starting point rather than a supported product — and remember HR data is sensitive, so vet what any server exposes.

It runs locally via Claude Desktop. You generate a Lattice API key, then add the server to your Desktop config:

{
  "mcpServers": {
    "lattice": {
      "command": "node",
      "args": ["/path/to/lattice-mcp/dist/index.js"],
      "env": { "LATTICE_API_KEY": "your-key" }
    }
  }
}

It needs Node.js 18+. This gives one person a working Claude + Lattice setup without touching org-level admin settings — at the cost of hosting and maintaining the server yourself.


Route 3: The Lattice public API + Claude Code

Under both MCP routes sits Lattice’s public REST API. An admin generates a key under Admin → Platform → API keys (it’s shown in full only once), and requests authenticate with a bearer token: Authorization: Bearer <key>. Note this is admin-gated — the key is tied to an active admin account, and if that person leaves, requests are rejected until a new key is issued. The surface is a fairly focused Talent API (users, goals, feedback, reviews, departments), not a sprawling one.

For a developer, the most direct path is skipping MCP entirely: give Claude Code the API key and let it write and run scripts against the endpoints — pull a review-cycle status, export goal progress, generate a headcount summary. It’s the most flexible route and the least point-and-click.

Worth noting: Lattice also has its own built-in AI for drafting reviews and summarizing engagement, so some of what you’d ask Claude to do already lives inside the product.


The limits that actually matter

Even with the official connector, the shape is “an analyst you query,” not “an agent that runs your people ops.” The limits that bite:

  • No triggers. MCP tools only run inside a session you start. There’s no “when a review cycle opens, nudge the managers who haven’t started” or “when a goal goes at-risk, alert the lead.” Nothing fires on a Lattice event — you have to be in the chat, prompting.
  • The local route needs Claude Desktop open. The community server runs on your own machine, so it only works while Claude Desktop is running. The official remote connector has no such constraint, and Claude’s scheduled tasks will run against it unattended — but they fire on a clock you set, not on a Lattice event. A review cycle is full of deadlines; none of them can be the trigger.
  • Admin- and region-gated. The official connector needs admin setup on both sides and is US-only at launch; the API key needs an active admin.
  • claude.ai on the web can’t reach a local server. The Desktop-config route is desktop-only; the web app can only use remote connectors your org has wired up.

So Claude is excellent for “help me understand where our review cycle and goals stand right now” and not built for “run the review cycle and chase the stragglers as things happen.”


If you want Lattice work that runs on its own: Carly

The moment you want people-ops work to happen without you in the chat — the instant a review cycle opens, the moment a goal slips — you’ve crossed past what a Claude connector is for.

Carly reaches Lattice through your own Lattice API key (paste it on the integrations page) — one of 200+ native integrations plus anything with an API. It’s built to act on triggers:

  • When a review cycle opens, Carly nudges the managers who haven’t started and compiles a status summary for HR — no one has to babysit the dashboard.
  • When a goal is created or updated, Carly posts a digest to your leadership Slack channel or emails a weekly at-risk-goals recap to people ops.
  • On a schedule, Carly pulls 1:1 gaps and stalled goals and sends a Monday summary — 24/7, in the cloud, and it sends the email itself rather than leaving it in a chat.

No-code setup. Tell Carly “I’d like managers nudged when a review cycle opens and a status compiled for HR” in plain English; it interviews you about the details, then builds the workflow with you. No prompt engineering, no server to host.

Actually sends. Carly drafts and sends email through Gmail and Outlook — so the nudge, the digest, the recap all go out on their own.

Connects to anything. 200+ native integrations plus any tool with an API via your own key, so Lattice changes can flow into email, calendar, tasks, and your CRM in one workflow.

AI agents start at $35/month, and steps in a workflow that don’t use AI run free and unlimited.


Claude vs Carly

Claude (Lattice MCP)Carly
Read/search goals, reviews, feedbackYesYes
Draft reviews in a sessionYesYes
Acts on triggers / eventsNoYes
Runs on a scheduleYes (scheduled tasks)Yes
Has its own inbox for work to land inNoYes
Nudges managers, compiles statusNoYes
Sends email as part of the flowNo (Gmail draft-only)Yes (Gmail + Outlook)
PricingPro $20 / Max $100–$200AI agents from $35/mo

Claude is the best way to reason over your Lattice data in a conversation; Carly is the way to run the review cycle and chase the stragglers on its own.


Frequently Asked Questions

Does Claude work with Lattice?

Yes. Lattice launched an official MCP server on July 7, 2026 at https://api.latticehq.com/mcp, added to Claude as a custom remote connector over OAuth. Once connected, Claude can read goals, reviews, feedback, and 1:1s and help draft reviews — using your existing Lattice login and permissions.

Is there an official Lattice MCP server in 2026?

Yes — this is one of the rare cases where the app ships its own. Lattice’s official MCP is a hosted remote connector for Claude, ChatGPT, and Slack, provisioned by an admin under Admin → Platform → MCP Server. It’s US data-residency only at launch, with EU support planned for fall 2026. A community server, EoinFalconer/lattice-mcp, also exists for self-hosting against the Lattice Talent API.

Can claude.ai on the web access Lattice?

Only through the official remote connector your admin has set up. The community server runs locally via Claude Desktop and can’t be reached by the web app. Either way, Claude looks at Lattice when it’s asked or when a timer says so — never because something in Lattice changed.

Can Claude run a review cycle or nudge managers automatically?

No. Every route only works inside a session you start — there are no event triggers. For automatic, trigger-based nudges and status compiling when a review cycle opens, you need an agent platform like Carly, which fires on events 24/7 in the cloud and sends the email itself. AI agents start at $35/month.


More: Claude connectors · Claude + BambooHR · Claude + Workday · Claude + Rippling · Can Claude send emails · Claude vs Carly

Ready to automate your busywork?

Carly schedules, researches, and briefs you—so you can focus on what matters.

See what people say

"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.

Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.

On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."

Gus Ibrahim, Founder & Director, IHR