What is the Squad AI remote MCP server?
The Squad AI remote MCP server is a hosted Model Context Protocol endpoint at https://mcp.meetsquad.ai/mcp, so AI assistants can connect to it without installing or running anything locally.
Squad AI MCP Server
@the-basilisk-ai
Basic information
Transport
stdio
Authentication
OAuth
Category
Memory & Knowledge
Author
the-basilisk-ai
Submitted by
Squad AI
About Squad AI MCP Server
Decision intelligence for product teams. Turn the noise of scattered feedback into signal you can build on with confidence.
Connection details
https://mcp.meetsquad.ai/mcpSetup
claude mcp add squad --transport http https://mcp.meetsquad.ai/mcpSquad AI MCP Server tools
30List every organisation and workspace you can access, with the current selection marked. Use it when the user mentions another workspace or a tool reports that selection is required.
Select which organisation and workspace subsequent tools operate on. Required when the user has access to more than one workspace.
One-call orientation for the current workspace: mission and description, top goals, recent signal activity, evidence-chain health, and open work counts. Call this first in a new session before other tools.
Update the current workspace's name, description, mission statement or logo. Returns the updated workspace.
People in the current organisation with their user IDs. Use it to resolve who to assign an action to — pass a returned userId as update_action's assignee.
Keyword search across signals, insights, actions, goals, documents and clusters by title/content match. Returns display IDs to pass to get_entity. Use list_* tools to browse instead when you have no search term.
Fetch any workspace entity by display ID (SI-1 signal, IN-1 insight, AC-1 action, GL-1 goal, BR-1 brief, DC-1 document, CL-1 cluster) or UUID. Documents return markdown. Pass include: ["evidence"] on insights for supporting signals with source links. Also the way to check on async work (brief generation, research).
Browse raw feedback signals with filters for source, type, sentiment, cluster and date range. Dismissed signals are excluded unless includeDismissed is set. Rows are summaries; use get_entity(SI-N) for full content.
Semantically related signals for a given signal (SI-N or UUID) — 'has anyone else said this?'. Useful before ingesting potentially duplicate feedback.
Browse signal clusters (recurring themes in feedback) with sizes and labels. Use get_cluster for members and linked insights.
A cluster's label, stats, member signals (capped) and linked insights. Pass the cluster UUID or CL-N from list_clusters.
Browse distilled insights ranked by combined score, with category/score/status filters or scoped to a goal. Use get_entity(IN-N, include: ["evidence"]) for the supporting signals.
Ranked actions (what the evidence says to do next) filtered by status, assignee, priority or parent insight. Defaults to open work: suggested and in_progress, snoozed excluded. Use get_action_context before working on one.
Everything needed to execute an action in one call: the action, why it exists (parent insight), the customer evidence behind it (top signals with source links), and the goals it serves. Call this before implementing or deciding on an action.
Strategic goals ordered by importance. Use get_entity(GL-N) for a goal's supporting insights.
The workspace change feed (humans and Squad agents), newest first — 'what happened since I last looked?'. Filter by entity type, actor type or agent; paginate with the cursor.
Connected feedback sources for this workspace and their sync health: provider, status, and each source's most recent sync time and total signals pulled. Use it to explain coverage gaps (e.g. no new signals in days) or to confirm a source is live before relying on it.
Pipe user feedback into the evidence chain (1–50 items). Content is deduplicated semantically server-side, then clustered and distilled into insights asynchronously. For a single item, consider find_similar_signals or search first to see if it's already known.
Move an action through its lifecycle: start (pick it up), complete (done — close the loop when work ships), dismiss (won't do), or snooze. Optionally append a note recording why, e.g. the PR that shipped it.
Edit an action's priority, effort, category or notes; assign it; or link it to an insight or brief. notes replaces the whole notes field (use update_action_status's note to append).
Create a strategic goal. Goals frame which insights matter; importance (1–5) drives ranking.
Update a goal's title, description or importance. A version snapshot is kept.
Curate an insight: set category or status, and link/unlink the goal it supports. Insight titles and descriptions are pipeline-generated and not editable.
Permanently remove a signal from the workspace (noise, spam, or mis-ingested content). This deletes it — it cannot be recovered.
Browse workspace knowledge documents (and briefs) with their paths and tags. Use get_entity(DC-N) to read one as markdown.
Create a knowledge document from markdown (research summaries, meeting notes, analyses). A document with a near-identical title returns the existing one instead — extend it with update_document. Don't start the markdown with the title as a heading; the title renders separately.
Replace a document's markdown body and/or title, and manage tags. A version snapshot is created, so the previous content stays recoverable in the app.
Briefs with their status (building/draft/in_review/finalised/failed) and recommendation. Read one with get_entity(BR-N).
Kick off AI generation of a brief from an action (AC-N) or insight (IN-N) — pass exactly one. Generation is asynchronous: poll the returned BR-N with get_entity until status moves past building. Pass retryBriefId instead to retry a failed generation.
Move a brief (BR-N) through its review lifecycle: draft, in_review, or finalised. Finalising records the decision. The building and failed states are managed by generation and can't be set here.
Overview
It's easier than ever to build, which means building in the wrong direction faster.
Your AI assistant can ship a feature in an afternoon. Product teams and their leaders need to judge whether that feature was worth building.
That answer is sitting in your feedback: spread across support tickets, sales calls, app store reviews, Slack threads, and analytics. Nobody has time to read it all, so the roadmap gets set by whoever spoke loudest in the last meeting.
Squad is the decision intelligence layer that fixes this. It finds the signal in the noise of your feedback and keeps it connected to the decisions it justifies, so you ship the right things.
How Squad earns its place
A team of always-on agents does the reading. They continuously research and read your feedback sources, identify signals and insights, and recommend actions that support your goals. By the time you ask a question, the evidence is already organised.
Every decision comes with receipts. Every action points back to the insight and the signals behind it. So "why is this on the roadmap?" has an answer you can show someone. A real customer quote, from a real source, with a link.
It remembers. Underneath is an evergreen knowledge base: a living memory of your product and company context that stays current as new evidence lands, with an auditable trail behind every change. Your assistant reads from it and adds to it, so it works from what your team already knows instead of a blank slate every session.
What this connector gives you
Squad, inside the assistant you already work in; Claude, ChatGPT, Cursor, or any MCP client. Instead of switching to a dashboard, you ask.
Ground a decision before you build it. Pull the exact customer signals behind any insight or action. Point your coding agent at a ticket and it gets the customer quotes, the insight that motivated the work, and the goal it serves — before it writes a line.
Triage feedback where you find it. Paste a support ticket, a call note, an angry tweet. It lands as a signal, deduped against existing themes, and Squad tells you whether it's a one-off or the eleventh instance of something you already know about.
Run your weekly review by asking for it. What changed, what the evidence says, what needs deciding, assembled from your data with receipts, not from spotty agent memory.
Get a decision brief on demand. One page from an action or insight: the recommendation, the evidence behind it, and the gaps worth knowing about.
Search everything at once. Semantic search across signals, insights, actions, goals, and documents.
Every entity has a short ID — SI-1, IN-42, AC-12 — so your assistant can cite the exact evidence behind every finding, and you can drill into any of it.
Try it
What's going on in my Squad workspace this week?
A customer said 'the CSV export times out on anything over 10k rows'. Tell me if it's a known theme.
What are people saying about exports? Show me the insight and the actual quotes behind it.
I'm about to pick up AC-31. Pull the customer evidence behind it first.
Draft a brief for AC-31 so we can decide whether to prioritise it.
Where the evidence comes from
Squad connects to the places your users already talk, and to the systems that record what you did about it:
| Customer voice | Gong calls, Slack channels, App Store, Play Store, Google Reviews, Trustpilot, Jira Service Management requests, Fellow meeting notes |
| Delivery | Linear, GitHub, Jira |
| Product & knowledge | PostHog, Notion |
Agents that run without being asked
Squad ships ~20 ready-made agents you can turn on in a click. A few of the ones teams keep:
- Morning briefing — what moved in the last 24 hours, citing the item behind every claim. Runs on day one, no setup.
- Escalation radar — reads your Slack channels daily for urgent customer language and unresolved questions, then ranks what needs attention.
- Build vs demand review — compares what shipped against what customers are actually asking for, and flags where delivery is running ahead of demand or real demand is going unbuilt.
- Competitor tracker — daily changes to your competitors' pricing, product, and positioning, kept as a maintained profile per competitor.
- Recurring bug detector — finds bug themes across your issue tracker and maintains a product-quality watchlist.
- Strategy consultant — a weekly review of goals, feedback, and market position, with candid recommendations and a log of past advice and what happened to it.
They report even on a quiet week — you get a written "nothing moved" rather than silence.
Get started
Claude Code:
claude mcp add --transport http squad https://mcp.meetsquad.ai/mcp
Claude Desktop: Settings → Connectors → Add Connector → https://mcp.meetsquad.ai/mcp
Everything else: point your client at https://mcp.meetsquad.ai/mcp. OAuth is discovered automatically.
You'll sign in with your Squad account in the browser on first use. Don't have one? Start at meetsquad.ai.
What's in the box
30 tools across workspaces, semantic search, signals, clusters, insights, actions, goals, briefs, documents, and feedback ingest — plus four one-shot workflows (triage-feedback, weekly-product-review, draft-decision-brief, ground-this-ticket) and two pinnable resources so strategy questions cost no tool calls at all.
Safe by default. Reads and writes are separated: writes need an explicitly granted scope, and every tool declares whether it's read-only. Access is isolated per user and per workspace and respects your existing Squad permissions.
Links
- Squad - meetsquad.ai
- Setup guide - docs.meetsquad.ai/mcp/overview
- Usage examples - link to GitHub
- Source - github.com/the-basilisk-ai/squad-mcp
- Privacy policy - meetsquad.ai/privacy-policy
- Support - support@meetsquad.ai
Frequently asked questions
How do I connect to the Squad AI MCP server?
Add the endpoint https://mcp.meetsquad.ai/mcp to any MCP-compatible client such as Claude Code, Cursor, or VS Code. The setup snippets on this page configure each client in one step.
Does the Squad AI MCP server require authentication?
Yes. Squad AI uses OAuth: the first time you connect, your MCP client opens a browser window to sign in and authorize access, then reuses the credentials for future sessions.
Which transport does the Squad AI MCP server use?
Squad AI exposes a stdio endpoint, the transport used by remote MCP servers and supported by all major MCP clients.
More remote MCP servers
A
AI Pricing Hub
Read-only, source-backed AI model pricing, history, comparisons and cost calculations.
A
AccountHub
One MCP connection for all your Gmail, Google Calendar, Drive and Contacts accounts, Slack workspaces and Notion — free.
Aard
Aard connects Macroeconomic and other official data from 170+ publishers including World Bank, IMF, BIS, ECB, Eurostat, and National Statistical Offices. Aard is powered by a large graph of the metadata universe and an
AdaptlyPost
AdaptlyPost: Schedule and publish social posts to Instagram, TikTok, YouTube, X, Facebook, LinkedIn, Pinterest, Threads, and Bluesky from any MCP client.
A
AI-Ready Webpage Extractor (Apify)
Convert web pages to clean Markdown and retrieval-ready chunks with browser rendering support.
Comments