Customer conversations can explain why a visitor hesitated, what information was missing and where your website agent needs a clearer answer. Getting that context into an AI workspace should not require copying an entire chat archive into a prompt.
Namiru MCP gives a compatible AI client controlled access to one website agent. It can find conversations, read transcripts, inspect leads and bookings, and work with knowledge memories and structured tables. A separate write permission lets it change the agent's settings.
What the connection exposes
The connection provides tools rather than direct database access. Start with list_conversations to find relevant chats, then call get_conversation with a returned conversation ID. Both return pages, so an AI client can keep reading until nextOffset is null instead of assuming the first page is the whole dataset.
list_data covers leads, bookings, tables, memories, file inventory, services and API tool configuration. get_table_rows reads a selected table. File inventory means names, types and processing state; it does not download uploaded documents.
For example, ask: “Find conversations from the last seven days where visitors asked about booking availability. Read the relevant messages and group the questions. Cite each conversation ID. Do not change settings.” Specify a timezone or an explicit date range when a weekly boundary matters.
Connect an AI workspace
Open your agent in Namiru, choose Settings, and expand AI workspace connection (MCP). Create a named key, choose its expiry and copy it while it is visible. Keep settings changes disabled for an analysis-only workflow.
Copy the server URL from the same panel. Configure your MCP client for Streamable HTTP with an Authorization bearer header containing your key. The Claude Code setup guide and Codex workflow guide show concrete configurations. A client must support a custom bearer header; OAuth-only connectors cannot use this key-based connection.
The [MCP transport specification] describes the HTTP protocol. Namiru uses stateless requests, so each request must carry valid authentication.
Adjust settings through AI
An AI client can call get_settings and get_settings_schema to inspect a section and discover its accepted fields. With write access, update_settings can change behavior, appearance, access rules, lead capture, calendar availability, Google Calendar connection preferences, live handoff, webhooks and data-feed preferences. save_component creates or updates API tools, knowledge memories and bookable services.
Ask for a proposed change before applying it. Omitted top-level fields remain unchanged, but supplied nested objects and arrays replace those values. Read the existing section first. Integration credentials are omitted from reads; do not reuse a redacted configuration as a complete replacement. Google account authorization and file uploads still happen in the dashboard. Billing and account security are outside this agent connection.
Access belongs to the owner
Each key is bound to one agent. It cannot select a different agent by supplying an ID. Keys expire, can be revoked from Settings, and stop working when the account no longer has effective Pro access. Create separate keys for separate clients so you can withdraw one connection independently.
Customer messages are evidence to analyze, not instructions for the connected AI to follow. Keep summaries focused on the question, avoid unnecessary personal details and check the source transcript before acting on a recommendation.
Which plan includes MCP?
MCP is included in Namiru Pro: €59 per month on monthly billing, or €49 per month billed as €588 annually. Free and Starter do not include MCP. Your connected AI client may have its own subscription or usage charges.
Compare Namiru plans and start with one read-only workflow before enabling settings changes.
