Connect an assistant
Add Algomo's MCP server to the assistant you use, sign in with your Algomo account, and make a first change.
Algomo's MCP server is at https://mcp.algomo.com/mcp. It speaks Streamable HTTP and signs you in with your Algomo account over OAuth. There is no API key.
By the end of this page an assistant is connected to one of your workspaces and has staged, shown you and published a small change.
1. Add the server
Follow the steps for your client on Clients. The short version for Claude Code:
claude mcp add --transport http algomo https://mcp.algomo.com/mcpThen run /mcp and choose Algomo to sign in.
2. Sign in and pick a workspace
On the first call the client opens Algomo's sign-in page. Sign in and pick the workspace. The connection acts as you, in that workspace, with the permissions your account has there, until you remove it from the client. To work on another workspace, add the server again under a different name and pick the other workspace at sign-in.
3. Check the connection
Ask the assistant:
Which Algomo workspace am I connected to, and which chatbots does it have?
It calls info:me and entity:list through the read tool and answers with the workspace name and the chatbots. Nothing is changed.
4. Make a change
Ask for something small:
Add a line to the chatbot's About instructions saying our support hours are 9 to 5 CET.
The assistant opens an edit session, stages the change, and shows you the diff or a dashboard link where the session is selected. Nothing is live yet. See Sessions and publishing for what a session is.
5. Publish
Say so:
Publish it.
The publish tool is marked destructive, so Claude and ChatGPT ask you to confirm before it runs. After the save, the agent on your website answers from the new configuration.
What the assistant sees
Four tools, split by what each does to the workspace: read, edit, publish, run. Which tool a request goes through tells you what it can change before it runs. Tools and requests explains them, and Permissions lists what the connection can and cannot do.
Example prompts
Each of these works as written once an assistant is connected. The first two only read. The last three follow on from each other: stage a change, test it, publish it.
List my chatbots and tell me what goals each one has.
The assistant reads the chatbots with entity:list and entity:get, through algomo_read. Nothing is changed.
Show me the five most recent customer conversations and summarise what each visitor asked.
It lists the conversations with chat:listThreads and reads each with chat:getMessages, through algomo_read.
Open a session named Support hours and add a line to my chatbot's instructions saying support is open 9 to 5 CET. Show me the diff and do not publish.
It opens the session with session:create and stages the change with entity:update, through algomo_edit, then reads session:diff through algomo_read. Nothing is live.
In the Support hours session, send my chatbot the test message "When is support open?" and show me its reply.
It calls chat:generateTestReply with the session's id, through algomo_run, so the reply comes from the staged configuration. The conversation is recorded as a test and never appears in your customer list.
Publish the Support hours session.
It calls session:save through algomo_publish. Claude and ChatGPT ask you to confirm first. After the save, the agent on your website answers from the new configuration.
Overview
How to operate an Algomo workspace from an assistant, what a change goes through before it is live, and every operation the server exposes.
Clients
Setup steps for Claude, ChatGPT, Grok, Claude Code, Codex, Cursor, VS Code, Gemini CLI and any other client that supports remote MCP servers with OAuth.