Available Tools

When you connect an MCP client to Sazabi, it can discover and use these tools. Each tool allows your AI assistant to interact with Sazabi in a specific way.

Tool reference

Project tools

list_projects

Returns all projects the authenticated user has access to. Use this to discover available projects before querying logs or starting conversations.

Parameters: None

Returns: Array of projects with:

• id: Project identifier
• name: Project name
• slug: URL-friendly identifier

Example prompt:

Use Sazabi to list my projects

Vercel tools

list_vercel_projects

Lists Vercel projects that are available for connection as data sources. Use this to discover which Vercel projects can be set up to send logs to Sazabi.

Parameters: None

Returns: Array of Vercel projects with:

• id: Vercel project identifier
• name: Project name
• framework: The framework used (e.g., Next.js, Remix)

Example prompt:

List my Vercel projects that I can connect to Sazabi

setup_vercel_data_source

Sets up a Vercel project as a data source for Sazabi. This configures the Vercel project to send logs and events to Sazabi for monitoring.

Parameters:

• vercelProjectId (required): The Vercel project ID to set up
• projectId (required): The Sazabi project to connect it to

Returns: Confirmation of the data source setup

Example prompt:

Connect my Vercel project to Sazabi

Thread tools

list_threads

Lists recent conversation threads in a project. Useful for finding existing investigations or continuing previous conversations.

Parameters:

• projectId (optional): The project to list threads from. If omitted, lists threads from the default project.
• limit (optional): Maximum number of threads to return
• cursor (optional): Pagination cursor

Returns: Array of threads with metadata

Example prompt:

Show me my recent Sazabi investigations in the backend project

search_threads

Performs full-text search across thread content. Find threads related to specific errors, services, or topics.

Parameters:

• query (required): Search query string

Returns: Array of matching threads

Example prompt:

Search Sazabi for threads mentioning "payment timeout"

get_thread

Retrieves a thread with all its messages. Use this to review the context of an investigation or understand what was discussed.

Parameters:

• threadId (required): The thread identifier

Returns: Thread object with:

• id: Thread identifier
• title: Thread title
• messages: Array of messages in the thread
• createdAt: When the thread was created
• updatedAt: When the thread was last updated

Example prompt:

Get the full conversation from my last Sazabi investigation

Message tools

send_message

Sends a message to Sazabi and receives a response. This is the primary way to ask questions about your systems through an external AI tool.

Parameters:

• projectId (required): The project context for the message
• threadId (optional): Continue an existing thread, or omit to start a new one
• message (required): The message content
• wait (optional): If false, return immediately and poll for response later. Defaults to true.
• timeoutSeconds (optional): Maximum time to wait for a response in sync mode. Defaults to 120.

Returns:

• In sync mode (wait: true): The response message
• In async mode (wait: false): A response ID to poll with get_response

Example prompt:

Ask Sazabi what errors occurred in the checkout service in the last hour

get_response

Polls for the response to a previously sent asynchronous message. Use this when you sent a message with wait: false and need to retrieve the response.

Parameters:

• responseId (required): The response ID from an async send_message call

Returns: The response message if ready, or a status indicating it is still processing

Example prompt:

Check if Sazabi has finished processing my last question

Sync vs async messages

The send_message tool supports two modes:

Synchronous (default)

When wait is true (the default), the tool waits for Sazabi to fully process the message and returns the complete response. You can adjust the timeout with timeoutSeconds. This is simpler but may take longer for complex queries.

Asynchronous

When wait is false, the tool returns immediately with a response ID. You can then poll with get_response to check when the response is ready. This is useful for:

• Long-running investigations
• When you want to ask multiple questions in parallel
• Avoiding timeouts on complex queries

Tool capabilities

Tools can do anything that Chat can do when you interact with Sazabi directly. This includes:

• Searching and analyzing logs
• Creating and updating threads
• Running investigations
• Generating summaries and reports

The main difference is that tools are invoked by your external AI assistant (Claude Code, Cursor, etc.) rather than through the Sazabi dashboard.

Next steps