Available Tekst MCP Tools – Tekst

This article lists the tools available through the Tekst MCP server. Your AI assistant uses these tools to retrieve and analyze data from your Tekst workspace, and to make changes such as building automations or maintaining AI models.

Tools that only read data are available to every workspace member. Tools that make changes require a workspace role with edit permissions - members with read-only access will not see them.

Users

List Users - Lists all members of the workspace, including their email addresses and roles. Returns summary fields only. Supports pagination.

AI models

List Models - Lists all AI models in your workspace (classification and extraction), including their names and statuses. Returns summary fields only - use Get Model for full information. Supports pagination.
Get Model - Returns detailed information about a specific AI model. For classification models, the response includes the model's labels (tags). For extraction models, it includes the model's entities. The modelKind field indicates which type was returned.
Create Model - Creates a new AI model (classification, extraction, or matching). The model starts empty - add tags or entities with the tools below. A new matching model cannot predict until records are imported, which is done from the Tekst app.
Update Model - Updates an existing model's name, description, or (for matching models) its primary key column and domain knowledge. Only the fields you provide are changed.
Add Classification Tag - Adds a tag to a classification model. Tags define the categories the model predicts and can be nested into a hierarchy (for example, "Billing > Refund Request").
Update Classification Tag - Updates an existing tag on a classification model. Only the fields you provide are changed. A tag can also be moved to the top level of the hierarchy.
Add Extraction Entity - Adds an entity (a structured data field to extract) to an extraction model. Entities can be nested to build structures such as individual line items under an order entity.
Update Extraction Entity - Updates an existing entity on an extraction model. Only the fields you provide are changed. An entity can also be moved to the top level of the structure.

Older generation trained models are read-only: the tools that modify models, tags, and entities only work on next-gen models.

Integrations

List Integrations - Lists all active data source integrations (such as Zendesk, Salesforce, Gmail, and others), including their names and types. Returns summary fields only - use Get Integration for full information. Supports pagination.
Get Integration - Returns configuration and sync status for a specific integration.
Update Integration - Updates an integration's properties, including renaming it, managing its group membership, and changing its sync state.

Threads and cases

These tools return raw message content, including email addresses, subject lines, and message bodies. They are only available when ticket display is enabled for your organization - an administrator can enable this in the organization settings.

Search Threads - Finds threads (conversations tied to an integration) and returns lightweight summaries with analytics metadata and classifications, without the raw message bodies. Search by free text, or look up directly by thread ID, message ID, sender, or recipient. Supports pagination.
Get Thread - Returns the full contents of a single thread, including sender and recipient email addresses, subject lines, and message bodies. Use Search Threads first to find the thread.
Get Case - Returns the full contents of a case within a process, including the underlying message content.

Process mining

List Processes - Lists all active process mining processes (discovered workflows such as ticket handling or order fulfillment), including their names, draft status, linked integrations, and throughput time target. Supports pagination.
Get Process Activities - Lists all activities (steps) discovered in a specific process. Activities can be classification-based (extracted from message content) or system-based (such as tag added or thread closed). Returns each activity's name, type, visibility status, occurrence count, and an optional description. Also indicates whether each activity was manually added - only manually added activities can be removed. Supports pagination.
Get Process Graph - Returns details for a specific process, including name, draft status, connected integrations, number of discovered variants, and throughput time target. Use Get Process Graph Paths to retrieve the actual variant paths with per-edge metrics.
Get Process Graph Paths - Returns individual paths (variants) through a process. Each path includes the sequence of activities, case frequency, and throughput time statistics per edge. Supports filtering by time range and integration, and sorting by case frequency, throughput time, or path length. Returns a link to view the variant in the Tekst UI when available.
Get Process Totals - Returns aggregate KPIs across all processes, including total case frequency, average throughput time, and the number of paths meeting the throughput time target. Supports time-range filtering.
Hide Process Activities - Hides one or more activities so they no longer appear in the process graph. Useful for cleaning up noise - the underlying data is kept and the change can be reversed with Unhide Process Activities.
Unhide Process Activities - Reactivates previously hidden activities so they appear in the process graph again.
Rename Process Activities - Renames one or more activities. Renaming several activities to the same name merges them into a single activity - the main way to consolidate duplicate or near-duplicate steps when cleaning up a graph.
Add Process Activities - Adds manual classification activities to a process, to track a step that was not automatically discovered.
Describe Process Activity - Sets or clears the description of a single activity, for example to clarify an ambiguous or auto-discovered step.

Analytics

Describe Analytics Tables - Returns the schema (column names, ClickHouse types, and descriptions) for all available analytics tables. Tables must always be prefixed with analytics. in queries (for example, analytics.mcp_threads). Always call this before running a query to confirm which columns and types exist. When aggregating data, always include a filter on thread_time, message_created_at, or day to limit the time range and avoid slow or timed-out queries.
Query Analytics - Executes a read-only SQL query against the ClickHouse analytics database. Tables must be prefixed with analytics. (for example, analytics.mcp_threads). Always include a filter on thread_time, message_created_at, or day when aggregating data - queries without a time filter may be slow or time out. Column metadata (types, constraints, and guardrails) is only available from Describe Analytics Tables - sampling rows does not expose this information. Use this to answer custom questions about your ticket data, message volumes, and automation performance.
Sample Analytics Table - Returns a sample of rows from one of the core analytics tables (mcp_threads, mcp_messages, or mcp_automation_block_runs) so you can inspect the data before writing a query. Sampling rows does not replace calling Describe Analytics Tables - always describe the schema before running a query.

Automations

List Automations - Lists all automation flows in your workspace, including their names and running status. Returns summary fields only - use Get Automation for full information. Supports pagination.
Get Automation - Returns full details for a specific automation flow, including its blocks (steps), creator, connected integrations, and aggregate success/failure stats over a recent time window.
Get Automation Monitor - Returns the execution history for a specific automation flow. Each entry represents a thread run with its status and block results. Supports filtering by block and status, with pagination.
Get Automation Run - Returns detailed information about a single automation run, including individual block results, the integration, and the thread involved. Use Get Automation Monitor first to find run IDs.
List Available Block Types - Lists all block types that can be used when building an automation flow. Use minimal detail for a quick overview (type, name, description, and branching shape) or full detail to also include configuration schemas. You can also filter to specific block types.
List Available Actions - Lists actions available for an automation flow or a specific integration, including any custom actions. Provide flow_id to list all actions for an existing automation (preferred), or integration_id to browse actions before creating one. Also supports filtering by search query. Use Get Action Schema to retrieve the full parameter details for any listed action.
Get Action Schema - Returns the parameter schema for a specific integration action or custom action, including what inputs are required, what values are valid, and what outputs the action produces. For integration actions, provide action_id along with flow_id or integration_id. For custom actions, provide app_id instead.
Get Condition Schema - Returns the condition group schema, valid variables, and operators for building condition blocks in an automation flow. Includes guidance on choosing AND vs. OR logic for condition groups.
Get App Code - Returns the source code of a custom action (app). Use Get Action Schema first - it reports the code's size, which is usually enough to decide whether reading the source is needed at all.

Building and editing automations

The tools below edit an automation's draft: changes only take effect once the automation is published from the Tekst app.

Create Automation - Creates a new automation flow with a name, connected integrations, and a trigger type. The automation starts without any processing blocks - use the block tools below to configure its steps.
Update Automation - Updates an automation's name, connected integrations, or trigger configuration.
Add Block - Adds a new processing block to an automation flow draft, with control over where it is placed and, for branching blocks such as conditions and approvals, which branch it follows.
Update Block - Updates an existing block's name, description, or configuration. Configuration changes are merged, so only the fields you provide are overwritten.
Remove Block - Removes a block from the draft. Blocks below the removed block become disconnected and can be rewired with Connect Blocks.
Connect Blocks - Connects two existing blocks by wiring the source block's output to the target block, with branch selection for branching blocks such as conditions and approvals.
Disconnect Blocks - Removes the connection between two blocks. The target block and everything below it become disconnected and can be rewired with Connect Blocks.
Unpublish Automation - Takes an automation offline so it stops running on incoming messages. The automation and its configuration are preserved, and it can be re-activated later.
Simulate Automation - Runs an automation against one existing thread and returns the outcome of every block. The default safe mode skips every action that changes state elsewhere (such as forwarding, replying, moving, or deleting), while other modes can perform some or all actions for real.
Decide Automation Approval - Approves, declines, or expires the approval step a run is paused on, then waits for the run to continue.