Autologyx MCP Work
Autologyx MCP Work is the tool collection that allows approved AI assistants and agents to help perform operational work in Autologyx.
It is called Work because it is intended for the people and teams who normally perform work through Autologyx: lawyers, attorneys, associates, legal operations teams, back office teams, external third parties, and in some scenarios clients. MCP Work gives those users a way to augment the work they already do in Autologyx by allowing approved agents to act alongside them, or in controlled autonomous workflows.
MCP Work is not for configuring the Autologyx product. It does not create Object Classes, design forms, build Sequences, configure permissions, or change administration settings. Instead, it focuses on live operational activity: finding records, reading and updating data, managing tasks, using document automation, navigating record relationships, and helping work move through an existing process.
Most users do not need to learn the MCP Work tool catalogue in detail before using an MCP-enabled agent. In normal use, the agent can read the available MCP tools, understand what each tool is for, and select the most appropriate tool for the user’s request. If you mainly want to understand how to begin using agents with Autologyx, start with Getting started with MCP.
Architects, administrators, solution designers, and advanced agentic workflow designers may want to understand the available tools in more detail. That is useful when planning permissions, designing autonomous patterns, assessing risk, or deciding which tools an agent should be allowed to use. The tool reference pages in this section explain the MCP Work tools in more technical detail.
What MCP Work is for
MCP Work is for agentic interaction with live Autologyx work.
Depending on the enabled tools, permissions, and customer configuration, an approved agent can help with activities such as:
- finding records that match a user request
- reading the details of a record
- updating structured fields on a record
- creating a new record
- creating related record structures
- navigating parent and child record relationships
- finding relevant task templates
- creating manual or automatic tasks
- assigning and unassigning tasks
- reading task details
- saving task data
- generating documents from configured templates
- checking document generation status
- downloading generated document payloads where permitted
This means an agent can do more than answer a question. It can help complete the operational steps that a user would otherwise perform directly in Autologyx.
Who benefits from MCP Work
MCP Work is designed for users and teams who interact with Autologyx as part of a live process.
This can include:
- lawyers and attorneys managing legal work
- associates preparing documents, triaging matters, or reviewing tasks
- legal operations teams managing intake, allocation, reporting, and process flow
- back office teams handling repeatable operational steps
- external third parties contributing to a controlled process
- clients or business users interacting with defined records and tasks
- autonomous service agents operating within tightly controlled permissions
The agent does not replace the Autologyx process. It works through that process using the tools it has been granted.
Why MCP Work matters
Many AI tools can summarise text or answer questions. MCP Work is different because it allows an approved agent to participate in the operational system of record.
For example, an agent might:
- review a new intake record,
- identify missing information,
- update structured fields,
- create a follow-up task,
- generate a first draft document,
- assign a review task to the correct team.
This can be useful in high-volume areas such as claims processing, client onboarding, remediation, contract management, legal intake, matter triage, and document-heavy workflows.
The value comes from combining agent reasoning with Autologyx control: structured data, record relationships, task management, document automation, permissions, and auditability.
What MCP Work is not
MCP Work is not a configuration toolkit.
It should not be used to:
- create or redesign Object Classes
- configure forms
- build or change Sequences
- create Sequence Actors
- configure roles or permissions
- change administration settings
- redesign the data model
Those activities belong to Autologyx configuration and administration workflows. MCP Work is for interacting with work that already exists inside a configured Autologyx environment.
For configuration guidance and platform knowledge, use Autologyx MCP Customer Success.
Co-pilot and autonomous patterns
MCP Work can support both co-pilot and autonomous patterns.
In a co-pilot pattern, the agent works alongside a named user. The user asks the agent to find information, prepare an update, draft a document, or create a task. The user remains closely involved and can review the output.
Some co-pilot style agents may occasionally ask the user to confirm an action before continuing. This is often seen where a request could result in a bulk operation or another higher-impact change. That confirmation behaviour is not triggered or controlled by Autologyx MCP Work tools. It is part of the safeguarding and guardrail behaviour implemented by the agent or AI client.
In an autonomous pattern, the agent may perform a defined set of actions without a user prompting each step. For example, it might monitor a queue of intake records, identify records missing key information, and create follow-up tasks.
Autonomous patterns require careful design. The agent should have narrow permissions, clear tool boundaries, and human review gates for important decisions or externally visible outputs.
Permissions and audit
MCP Work should be enabled with governance in mind.
Agent access can be controlled by the tools exposed to the agent, the account the agent uses, the user context, and the underlying Autologyx role-based access model. Customers can decide which agents can read data, update records, create tasks, generate documents, or perform other permitted actions.
Audit logging is especially important because MCP Work can include write actions. Customers should be able to review which tools were called, what action was requested, what data was affected, when the action happened, and whether the action succeeded.
Record tool endpoints
Record tools allow an approved agent to find, read, create, update, relate, and work with Object Records and record-related documents.
For more detail, see Available MCP Work tools and Available MCP Work tools.
| Endpoint | What it does | Example use case |
|---|---|---|
check_object_records_hierarchy_creation | Checks the status of an asynchronous record hierarchy creation process. | An agent confirms whether a complex matter structure has finished creating. |
create_object_record | Creates a new Object Record with supplied field values. | An agent creates a new intake record from information provided by a user. |
create_object_records_hierarchy | Creates a parent Object Record and nested child Object Records as an asynchronous operation. | An agent creates a matter structure with related parties, requests, or child items in one operation. |
download_object_record_document | Downloads a generated Object Record document payload and metadata where permitted. | An agent retrieves a completed document for review or onward processing. |
generate_object_record_document | Starts document generation for an Object Record and document template. | An agent generates a first draft document from record data. |
get_document_template | Retrieves details for a specific document template. | An agent checks template details before generating a document. |
get_object_class | Retrieves details for a specific Object Class. | An agent checks the structure of a matter type before explaining what record data is available. |
get_object_record | Retrieves details for a specific Object Record. | An agent reviews the current state of a matter before suggesting next steps. |
get_object_record_document | Retrieves document generation metadata, processing status, and download information. | An agent checks whether a generated document is ready. |
list_object_class_children | Lists direct child Object Classes for a parent Object Class. | An agent confirms which related child records can be created under a parent matter. |
list_object_class_fields | Lists valid fields and aliases for a specific Object Class. | An agent identifies the correct field aliases before creating or updating a record. |
list_object_class_parents | Lists direct parent Object Classes for a child Object Class. | An agent works out which parent records a child record can be related to. |
remove_object_record_child | Removes the relationship between a parent record and a child record. | An agent disconnects a record that was linked to the wrong parent, where permitted. |
search_document_templates | Searches document templates for a specific Object Class. | An agent finds the correct template for an engagement letter or approval document. |
search_object_classes | Searches available Object Classes. This supports discovery of the data model without changing configuration. | An agent finds the correct Object Class before searching for matter records. |
search_object_record_children | Finds child records linked to a parent Object Record. | An agent lists all related contract documents or subsidiary requests under a matter. |
search_object_record_documents | Searches documents linked to an Object Record. | An agent finds documents already generated or attached to a matter. |
search_object_record_parents | Finds parent records linked to a child Object Record. | An agent identifies the matter or client record linked to a document record. |
search_object_records | Searches Object Records using filters, pagination, ordering, and selected fields. | An agent finds all open intake records for a client or process. |
update_object_record | Updates fields on an existing Object Record. | An agent fills missing triage fields after reviewing source information. |
Task tool endpoints
Task tools allow an approved agent to find, create, read, update, assign, unassign, and manage Tasks.
For more detail, see Available MCP Work tools and Available MCP Work tools.
| Endpoint | What it does | Example use case |
|---|---|---|
assign_task_to_user_groups | Assigns a Task to one or more user groups. | An agent assigns an intake task to the legal operations team. |
assign_task_to_users | Assigns a Task to one or more users. | An agent assigns a contract review task to the responsible lawyer. |
create_automatic_task | Creates a new Automatic Task from a Task Template and Object Record. | An agent starts a configured task as part of a repeatable process. |
create_manual_task | Creates a new Manual Task against an Object Record. | An agent creates a lawyer review task with instructions and a due date. |
get_task | Retrieves current details, schema, and response data for a Task. | An agent reviews what information is needed before updating or saving a task. |
get_task_template | Retrieves details for a specific Task Template. | An agent checks what task type should be used for legal review. |
save_task | Saves or updates Task data while keeping the task outstanding. | An agent records partial progress on a task that still requires human action. |
search_task_templates | Searches available Task Templates. | An agent finds the right template before creating a follow-up task. |
search_tasks | Searches Tasks using filters, pagination, ordering, and assignment criteria. | An agent finds overdue tasks assigned to a team or user. |
unassign_task_from_user_groups | Removes one or more user group assignments from a Task. | An agent removes a team assignment after the task is reassigned to a named user. |
unassign_task_from_users | Removes one or more user assignments from a Task. | An agent removes a user assignment when work is reallocated. |
How the tools work together
The real value of MCP Work comes from combining tools.
For example, an agent might:
- search for the right intake record,
- retrieve the record details,
- inspect related child records,
- update missing structured fields,
- search for a task template,
- create a manual task,
- assign the task to a user group,
- generate a document from a configured template.
Each step is performed through a tool that can be permissioned, controlled, and audited.
For more information, see Available MCP Work tools.
Designing safe MCP Work patterns
MCP Work gives agents operational capability, so design matters.
Before enabling an agent, consider:
- which users or service accounts the agent will act with
- which Object Classes the agent can access
- which records the agent can read or update
- which task templates the agent can use
- whether the agent can create or complete tasks
- whether the agent can generate or download documents
- which actions require human review
- how tool calls will be audited
- how quickly access can be removed if needed
Start narrow. Give the agent the smallest set of tools and permissions needed for the use case, then expand only where there is a clear benefit.
Important reminder about agentic work
Important reminder
MCP Work can allow agents to perform real actions in Autologyx.
Even where permissions, tools, prompts, and workflows have been carefully designed, an agent may misunderstand a request, choose the wrong tool, rely on incomplete context, or produce an unexpected result.
Permissions, audit logging, validation, deterministic checks, and human review can reduce risk, but they cannot remove it entirely.
Use human review where an agent action affects important decisions, external communications, sensitive data, record updates, document generation, or task completion.
Things to remember
- MCP Work is for operational work in Autologyx, not product configuration.
- It is designed for the people and teams who already work through Autologyx processes.
- MCP Work can support both co-pilot and autonomous agent patterns.
- Most users do not need to learn each MCP tool in detail because agents can navigate the available tool catalogue.
- Advanced designers should understand the tool catalogue when planning permissions, autonomous workflows, and risk controls.
- Record tools allow agents to search, read, create, update, relate, and work with Object Records and documents.
- Task tools allow agents to search, create, assign, unassign, save, and manage Tasks.
- Tool access should be limited to the tools and data the agent genuinely needs.
- Write access should be granted carefully and tested before production use.
- Audit logging and human review are essential for higher-risk workflows.