Connect multiple MCP servers, bring their tools together, and chain calls into new tools. KNWN Compose publishes the result as a new MCP server your AI can connect to.
GitHub, Linear, and Slack tools exposed through one MCP server.
publish_release_updateCompose gives you two ways to build a new MCP server. Use either one, or use both together.
Publish several focused tools backed by different apps on one MCP server.
get_customer_contextopen_support_ticketshare_account_contextCustomer Ops MCPYour AI connects to one server and sees only the composed tools you publish.
Turn several MCP calls into one higher-level tool.
Intercom.get_conversationStripe.get_customerSlack.post_messageCustomerOps.share_account_contextYour AI calls one tool. Compose handles the calls behind it.
Four steps take you from the servers you already connect to, to one endpoint your AI can call.
Start composing Try the zero-credential demoAdd an eligible hosted server, or an approved remote MCP server in a design-partner workspace. Compose discovers the tools, inputs, outputs, and permissions each server exposes.
Choose the app accounts a tool may use, then chain their source actions into one focused capability.
Give the server a name. Choose its tools. Define the inputs and results that MCP clients will see. The source servers stay behind the composition.
Call the tools with test inputs. Inspect every source call and result. Then publish the approved version at a stable MCP endpoint.
Connect eligible official hosted servers. Approved design-partner workspaces may also connect compatible remote MCP endpoints. Their tools become building blocks for the MCP servers you create.
Add a connectionUse repository, issue, pull request, and GitHub automation tools in your new MCP server.
Use tools from your Linear workspace through Linear's hosted MCP server.
Use tools from your Slack workspace through Slack's hosted MCP server.
Use pages, databases, and other tools from your Notion workspace.
Use Jira issue and Confluence tools through Atlassian's remote MCP server.
Use issue, error, release, and project tools from your Sentry organization.
Use contacts, companies, deals, and engagement tools from HubSpot.
Use customer, subscription, invoice, and payment tools from Stripe.
Use conversation, contact, and help center tools from Intercom.
Approved organizations can connect a compatible remote MCP server after platform endpoint checks.
An AI should not have to understand several MCP servers and reconstruct the same sequence of calls every time. Compose moves that structure into a purpose-built MCP server.
KNWN Builder can suggest source tools, call order, input mappings, and output schemas. It cannot access credentials, call connected servers, publish changes, or alter a live server. Published MCP servers run the reviewed composition. A language model does not decide what to call next during execution.
Build inside the boundaryStart with the MCP servers you already use. Combine the right tools, add the chains you need, and publish one clean MCP endpoint.
Create an MCP server Run the live demo