Skip to content

Connect Actionstep to ChatGPT: Manage Matters, Billing, and Files

Learn how to connect Actionstep to ChatGPT using an MCP server. Automate legal matters, time entries, and billing with our step-by-step integration guide.

Sidharth Verma Sidharth Verma · · 9 min read
Connect Actionstep to ChatGPT: Manage Matters, Billing, and Files

If you need to connect Actionstep to ChatGPT to automate legal practice workflows, orchestrate matter management, or log billable time using natural language, you need a Model Context Protocol (MCP) server. This server acts as a standardized translation layer between ChatGPT's JSON-RPC tool calls and Actionstep's highly specialized REST APIs.

If your team uses Claude, check out our guide on connecting Actionstep to Claude or explore our broader architectural overview on connecting Actionstep to AI Agents.

Giving a Large Language Model (LLM) read and write access to a comprehensive practice management system like Actionstep is a major engineering undertaking. You either spend weeks building, hosting, and maintaining a custom MCP server to translate LLM arguments into Actionstep's specific JSON structures, or you use a managed infrastructure layer to generate tools dynamically.

This guide breaks down exactly how to use Truto to generate a secure, authenticated MCP server for Actionstep, connect it natively to ChatGPT, and execute complex legal workflows using natural language.

The Engineering Reality of the Actionstep API

Building a custom MCP server means you own the entire API lifecycle. While the open MCP standard provides a predictable way for models to discover tools, implementing it against Actionstep's specific data model is exceptionally painful.

If you decide to build a custom Actionstep MCP integration in-house, here are the specific API challenges you will have to engineer around:

Actionstep follows a specific JSON API convention where payloads are tightly wrapped. When you want to update a task or create a time entry, you cannot just send a flat JSON object. The API requires the fields to be wrapped in a root object matching the resource name (e.g., {"tasks": {...}}). Furthermore, related records - like associating a time entry with a specific matter or participant - must be explicitly defined under a nested links object. Without dynamic schema generation, an LLM will constantly hallucinate flat JSON structures and fail 400 Bad Request validations.

File Chunking Complexity

Document management is critical in legal workflows, but Actionstep does not allow simple multipart file uploads for large documents. The Actionstep /files endpoint requires developers to upload files in exact 5 MB chunks, specifying a part_number and part_count for every request. The first chunk returns an ID that must be sequentially passed into subsequent requests. Training an LLM to natively orchestrate this multi-step chunking logic is nearly impossible without an intermediary abstraction layer.

Action Types and Matter Creation

In Actionstep, a "Matter" is referred to as an "Action" under the hood. You cannot simply POST /actions. To create a new matter, you must first interact with the actioncreate endpoint and supply a specific action_type_id. This dictates what required fields, custom data collections, and participant roles are necessary to instantiate the matter. Your custom MCP server would need to dynamically read these requirements and expose them to ChatGPT as strict JSON schemas at runtime.

How to Generate an Actionstep MCP Server

Instead of building a schema translator from scratch, Truto dynamically generates an Actionstep MCP server by deriving tool definitions directly from the integration's documented endpoints.

You can generate your Actionstep MCP server in two ways: via the Truto dashboard or programmatically via the API.

Method 1: Via the Truto UI

  1. Log into your Truto account and navigate to Integrated Accounts.
  2. Select your connected Actionstep account (or connect a new one using the OAuth flow).
  3. Click on the MCP Servers tab.
  4. Click Create MCP Server.
  5. Configure your server by giving it a name and selecting specific allowed methods or tags (e.g., restricting the LLM to only "read" operations).
  6. Click Create and copy the generated MCP Server URL.

Method 2: Via the Truto API

For teams building automated deployment pipelines, you can generate an MCP server using a single API call. Truto generates a secure, hashed token stored in distributed KV storage to authorize requests.

curl -X POST https://api.truto.one/integrated-account/<YOUR_INTEGRATED_ACCOUNT_ID>/mcp \
  -H "Authorization: Bearer $TRUTO_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Actionstep for ChatGPT",
    "config": {
      "methods": ["read", "write"],
      "tags": ["actions", "billing", "tasks"]
    }
  }'

The API returns a payload containing a single, secure url. This URL acts as the routing and authentication layer for the MCP protocol.

{
  "id": "mcp-token-789",
  "name": "Actionstep for ChatGPT",
  "config": { "methods": ["read", "write"] },
  "expires_at": null,
  "url": "https://api.truto.one/mcp/a1b2c3d4e5f6..."
}

How to Connect the Actionstep MCP Server to ChatGPT

Once you have your Truto MCP server URL, you must register it with your ChatGPT environment. You can do this through the ChatGPT UI for cloud-based interactions or via a manual configuration file for local/custom setups.

Method 1: Via the ChatGPT UI

Note: Custom MCP connectors currently require a ChatGPT Pro, Plus, Business, Enterprise, or Education account with Developer mode enabled.

  1. Open ChatGPT and navigate to Settings -> Apps -> Advanced settings.
  2. Toggle Developer mode to ON.
  3. Under MCP servers / Custom connectors, click to add a new connector.
  4. Name: Enter a recognizable name (e.g., "Actionstep (Truto)").
  5. Server URL: Paste the https://api.truto.one/mcp/<token> URL you generated in the previous step.
  6. Click Save.

ChatGPT will immediately send an initialize JSON-RPC handshake to the server and discover the available Actionstep tools.

Method 2: Via Manual Config File

If you are running a local agent wrapper or interacting with ChatGPT via a desktop client that relies on local config files (similar to Claude Desktop), you can connect via a Server-Sent Events (SSE) proxy command.

Create or update your MCP configuration JSON file:

{
  "mcpServers": {
    "actionstep_truto": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-sse",
        "--url",
        "https://api.truto.one/mcp/<YOUR_SECURE_TOKEN>"
      ]
    }
  }
}

This configuration instructs the client to proxy standard stdio MCP communication over HTTP POST to Truto's remote endpoint.

Actionstep Hero Tools for ChatGPT

Truto automatically maps Actionstep endpoints to descriptively named MCP tools. Here are six high-leverage hero tools your AI agents can use to manage your practice.

list_all_actionstep_actions

Returns a paginated collection of Matters (Actions) across the firm. This tool allows the LLM to search for active cases, retrieve matter IDs, and filter by matter status.

"Find the latest open matters assigned to our real estate division. List their names, creation dates, and the primary matter ID for each."

create_a_actionstep_time_entry

Logs billable time against a specific matter. The LLM handles the required Actionstep formatting, supplying the timestampDate, actualHours, and associating the time entry via the links object to the correct matter ID.

"Log 2.5 hours of billable time for today on matter ID 142. The description should read: 'Review and redline commercial lease agreement.'"

list_all_actionstep_bills

Retrieves a collection of invoices (bills) associated with specific matters. The agent can paginate through this data to audit unpaid balances or summarize monthly billing.

"Check the billing history for matter ID 890. Are there any outstanding bills from the last 60 days? If so, summarize the totals."

update_a_actionstep_task_by_id

Updates the properties of a specific task within Actionstep. This is frequently used to progress workflows, change task statuses, or update due dates.

"Find task ID 5512 and update its status to 'Complete'. Then, push the due date of task ID 5513 to next Friday."

list_all_actionstep_action_documents

Queries the metadata of files and documents associated with a matter. This allows the LLM to verify if critical case files have been uploaded before progressing a workflow.

"List all the documents currently stored under matter ID 305. I need to check if the 'Final Settlement PDF' is in the folder."

create_a_actionstep_participant

Creates a new global contact or organization in Actionstep. The LLM can extract contact details from an email or summary and insert them directly into the firm's directory.

"Take the contact details from the email I just pasted and create a new participant in Actionstep. The display name is 'Jane Doe', and the email is 'jane@example.com'."

To view the complete inventory of available tools, including detailed query schemas and required body parameters for each, visit the Actionstep integration page.

Workflows in Action

By chaining these tools together, ChatGPT can act as an autonomous legal assistant, orchestrating complex operations across Actionstep's API.

Scenario 1: Matter Triage and Task Delegation

When a new client reaches out, a partner needs to quickly identify the active matter, check for outstanding documentation, and assign a review task to a paralegal.

"Find the active matter for 'Smith Real Estate Merger'. Check if the 'Term Sheet' document is uploaded. If it is, create a new task to review the document by tomorrow."

  1. Search Matters: The agent calls list_all_actionstep_actions with a filter query to locate the matter ID for "Smith Real Estate Merger".
  2. Verify Documents: The agent extracts the matter ID and calls list_all_actionstep_action_documents, analyzing the response to verify the "Term Sheet" exists.
  3. Create Task: The agent calls create_a_actionstep_task, passing the required name, status, dueTimestamp, and linking it to the matter ID.
sequenceDiagram
    participant User as User
    participant ChatGPT as ChatGPT (MCP Client)
    participant Truto as Truto MCP Server
    participant Actionstep as Actionstep API

    User->>ChatGPT: "Find Smith matter, check docs, assign task."
    ChatGPT->>Truto: tools/call (list_all_actionstep_actions)
    Truto->>Actionstep: GET /actions?name=Smith
    Actionstep-->>Truto: 200 OK (Returns Action ID 55)
    Truto-->>ChatGPT: Tool Result
    ChatGPT->>Truto: tools/call (list_all_actionstep_action_documents)
    Truto->>Actionstep: GET /actiondocuments?action_id=55
    Actionstep-->>Truto: 200 OK (Term Sheet found)
    Truto-->>ChatGPT: Tool Result
    ChatGPT->>Truto: tools/call (create_a_actionstep_task)
    Truto->>Actionstep: POST /tasks
    Actionstep-->>Truto: 201 Created
    Truto-->>ChatGPT: Tool Result
    ChatGPT-->>User: "Task created for Term Sheet review on matter 55."

Scenario 2: Billing Reconciliation

At the end of the week, a senior associate needs to ensure their time is logged and check the status of a specific client's invoices.

"Log 3 hours of time for today on matter ID 210 with the description 'Deposition prep'. Then, list all bills for that matter and tell me which ones are unpaid."

  1. Log Time: The agent immediately calls create_a_actionstep_time_entry with actualHours: 3 and the description, linking it to ID 210.
  2. Fetch Invoices: The agent calls list_all_actionstep_bills filtered by the matter ID.
  3. Analyze & Summarize: The agent reads the JSON response, filtering out paid bills in its context window, and formats a summary for the user.

Security and Access Control

Handing an LLM unrestricted access to your firm's Actionstep environment is a massive security risk. Truto's MCP architecture provides granular access controls at the server level, ensuring the AI can only touch what you explicitly allow.

  • Method Filtering: When creating the server, you can restrict operations to ["read"] only. This strips out all create, update, and delete tools, ensuring the agent can query case files but cannot alter billing data.
  • Tag Filtering: Limit the server's scope by resource domains. Passing tags: ["billing", "time"] ensures the server only exposes tools related to financials, hiding sensitive participant data or document contents.
  • Require API Token Auth: By setting require_api_token_auth: true, the MCP server URL is no longer sufficient on its own. The client must also pass a valid Truto API token in the Authorization header, adding a critical second layer of identity verification.
  • Time-to-Live (TTL): You can supply an expires_at ISO datetime when generating the server. Once the timestamp is reached, distributed cleanup alarms automatically destroy the token in the KV store, instantly revoking the LLM's access.

Handling Rate Limits in Production

Actionstep enforces strict rate limits to protect their infrastructure, typically returning HTTP 429 Too Many Requests when limits are exceeded.

It is important to note that Truto does not retry, throttle, or apply backoff on rate limit errors. If Actionstep returns a 429, Truto passes that error directly back to the caller. However, Truto normalizes the upstream rate limit information into standardized HTTP headers (ratelimit-limit, ratelimit-remaining, ratelimit-reset) per the IETF specification.

This ensures that your agent framework (like LangGraph or AutoGen) can reliably parse the reset window and apply its own retry or backoff logic without having to understand Actionstep's proprietary header formats.

Connect Actionstep to ChatGPT Today

Connecting ChatGPT to Actionstep doesn't have to mean maintaining thousands of lines of fragile API integration code. By leveraging Truto's auto-generated MCP servers, you can instantly give your AI agents secure, schema-aware tools to interact with matters, billing, and documents.

Stop wrangling Actionstep's nested links objects and file chunking logic. Generate your tools dynamically and focus on building high-leverage legal workflows.

FAQ

Does Truto automatically retry Actionstep API requests if they hit rate limits?
No. Truto passes HTTP 429 rate limit errors directly back to the caller. However, it normalizes the rate limit information into standard headers (ratelimit-limit, ratelimit-remaining, ratelimit-reset) so your AI agent framework can handle the retry and backoff logic predictably.
Can I prevent ChatGPT from modifying Actionstep data?
Yes. When generating the MCP server in Truto, you can use method filtering (e.g., config.methods: ["read"]). This prevents any create, update, or delete tools from being exposed to the LLM.
Do I need to build custom schemas for Actionstep's nested objects?
No. Truto automatically generates the required JSON-RPC tool schemas dynamically based on Actionstep's API documentation, handling complex nested relationships like the 'links' object natively.
What ChatGPT plan do I need to use custom MCP connectors?
To connect remote MCP servers like Truto via the UI, you need a ChatGPT Pro, Plus, Business, Enterprise, or Education account with Developer mode enabled.

More from our Blog