Connect your docs and your business data to Claude, Cursor & AI agents.
One MCP endpoint gives any agent that speaks the Model Context Protocol read-only access to your team's docs, what your team knows across 40+ connected apps, and your own code — so Claude and Cursor can answer "who signed up this week" and "who owes us money", not just "what does the doc say".
Start free — connect an agentIncluded in the free plan · no credit card required
What an MCP endpoint unlocks
Agents read your docs and your data
Cursor, Claude and your own agents pull your team's docs, knowledge and code when they answer.
One standard, no custom glue
MCP is the open protocol AI tools support — connect once, no per-tool integrations.
Answers grounded in your business
Agents pull from your real docs, tools and numbers instead of guessing, so answers match your team's truth.
What you can actually ask
Seven read-only tools across 40+ connected apps. Your workspace shows exactly which of these are live for you — and what to connect for the rest — on the MCP page in the app.
What does the team already know?
Search by topic across the tools that hold prose — not by customer.
- “What's our refund policy?”
- “Why did we drop the Postgres migration?”
- “What do we tell customers who ask about data residency?”
Reads from: Slack public channels, Notion page contents, your help centre, saved replies
What's written down?
Your private knowledge base, listed and read in full — the docs your team keeps, available to the agent your team already uses.
- “Summarise our onboarding doc.”
- “What does the incident runbook say to do first?”
- “Which docs mention the Postgres migration?”
Reads from: Your Ziikly knowledge base — no integration required
How is this implemented?
Search and read your connected GitHub repositories, so an agent answers from the code rather than from memory.
- “Where do we handle Stripe webhooks?”
- “Show me the refresh-token logic.”
- “Is there a Fortnox integration in the codebase?”
Reads from: Your connected GitHub repositories
What do we even sell?
Your own product names, plan prices, deal stages, courses and forms — so an agent uses your vocabulary instead of guessing at it.
- “What plans do we sell and what do they cost?”
- “What are the stages in our sales pipeline?”
- “Which onboarding forms do we have live?”
Reads from: Stripe, Paddle, Shopify, Chargebee, Recurly, HubSpot, Pipedrive, Thinkific and more
Read-only, and narrower than you'd expect
Every connection uses restricted, read-only, encrypted-at-rest credentials: Ziikly can display your data but can never move money, send mail or change anything. Some limits go further than read-only, on purpose:
- Slack: public channels only. Topic search never reads DMs or private channels. If a match was in one, you're told it was skipped rather than left to assume the answer was complete.
- Help centres, not tickets. Topic search reads your published articles. A customer's actual support thread is only reachable once you've named that customer.
- Internal notes stay labelled. When an agent reads a ticket, private agent notes are clearly marked so they never get quoted back to a customer.
- Customer data is never warehoused. It's fetched live from each tool on every question and thrown away after.
How the MCP endpoint works
- 1.Your team writes docs in Ziikly and connects the tools you already pay for — Stripe, HubSpot, Slack, Fortnox, whatever you run on — with read-only keys.
- 2.You point your AI agent at Ziikly's MCP endpoint. One connection, not one per tool.
- 3.The agent reads what it needs to answer — docs, a customer's history, this week's leads, the overdue list. Private to your team, read-only, live on every question.
Where teams would use it
Coding assistants
Cursor pulls your architecture docs before writing code against your stack.
Claude workflows
Claude reads your process docs to draft SOPs, emails and plans your way.
Onboarding a hire
They ask the agent instead of interrupting the team — it answers from your docs and your Slack.
Support answers
An agent checks your help-centre articles and saved replies before anyone drafts a response.
Frequently asked questions
What can an AI agent actually ask for?+
Seven read-only tools: ask what your own workspace is connected to; list and read your team's docs; search what the team knows by topic across Slack public channels, Notion, your help centre and saved replies; list your own products, plans, pipeline stages, courses and forms; and search and read your connected GitHub repositories. Customer records are deliberately not among them — agents get your team's own material, never your customers' data. Inside the app, the MCP page shows exactly which tools are live for your workspace and what to connect for the rest.
Can an agent read our Slack DMs or private channels?+
No. Topic search covers public channels only, and that limit is enforced in our code rather than left to the token's permissions. If a match was in a DM or private channel, the answer tells you it was skipped instead of quietly presenting a partial result as complete. Support tickets work the same way: topic search reads your published help-centre articles, never the conversations themselves — an agent has no route to a customer's correspondence at all.
Is it really read-only?+
Yes. Every connection uses restricted, read-only, encrypted-at-rest credentials, and every tool is a GET or a search. Ziikly can display your data but can never move money, send mail, or change anything in a connected tool. Where a vendor offers no read-only token — Asana and Monday.com issue full-access user tokens — we say so in the setup screen rather than letting you assume otherwise.
What is an MCP endpoint?+
MCP (Model Context Protocol) is the open standard that lets AI tools — Claude, Cursor, and other agents — connect to your data. An MCP endpoint for your knowledge base means an AI agent can read your team's docs directly, through one standard connection.
Which tools work with MCP?+
MCP is supported by Claude, Cursor, and a growing list of AI agents and coding assistants. Any tool that speaks MCP can connect to your Ziikly knowledge base.
Is this the same as RAG / AI search?+
Related but different. AI search (Ask AI) answers questions inside Ziikly with your docs, team knowledge and code as the source — a Pro feature that's free right now. An MCP endpoint exposes that same material to outside AI agents (Claude, Cursor) so they can pull your docs when helping your team. Neither one can read your customer records.
Is the MCP endpoint available now?+
Yes — it's live today on every plan, including free. Sign up, write a doc, then open the MCP endpoint page in the app to get your endpoint URL and access token.
Is the MCP endpoint included in the free plan?+
Yes. The free plan includes the private knowledge base and the MCP endpoint — unlimited docs and team members. Connect Claude, Cursor or any MCP client to your team's docs. In-app AI search (Ask AI) is a Pro feature that's free for every workspace right now.
Related ways to use Ziikly
Free knowledge base
The private, Markdown-based docs your MCP endpoint would expose.
AI search & RAG
Ask questions and get answers grounded in your docs, with cited sources.
Connect AI to your knowledge base
The plain-English guide to hooking an AI assistant up to your team docs.
Read-only customer dashboards
The module that gives customers and teams a zero-write view of live data.
Point your agent at your knowledge base.
The MCP endpoint is live today on every plan. Sign up free, connect a tool or two, then grab your endpoint + token from the MCP page in the app and paste them into Claude, Cursor or any MCP client. That page also lists exactly what your workspace can be asked for, and what to connect next.
Start free — no card requiredMCP endpoint included in the free plan · Ask AI free now, Pro at launch.