AI assistants

Your team can work on warranty and vehicles from an AI assistant in plain language. For example: which submitted claims for a dealership are still waiting, what a claim contains, adding a labour line to a draft, or running the next lifecycle step. Fleetlane offers this through a hosted server that speaks the Model Context Protocol (MCP), the open standard AI assistants use to connect to business systems. Any assistant that supports remote MCP servers can connect, either by signing in or with a personal API key.

It acts as the person using it

The assistant has no access of its own. Every request it makes is handled as the person who connected it:

• Same roles and permissions. It sees and changes only what that person can in Fleetlane.
• Same dealership scope. A dealer’s assistant sees that dealership’s claims and nobody else’s.
• Same field rules. Hidden fields stay hidden, and fields the person cannot edit in the record’s current status are refused.
• Same lifecycle steps. It can run only the steps the person’s role may run at that moment.

Each tool calls the same procedure the screen uses, so the checks and refusals are the app’s own, and every change is recorded as that person. There is no separate data feed and no broad service account. When the person’s access ends, so does the assistant’s. A connection for one organisation cannot be used against another.

Connect an assistant

Your MCP address is your workspace address followed by /mcp, for example https://<workspace-address>/mcp.

1. Add that address to your assistant as a remote MCP server (in Claude, a custom connector).
2. The assistant finds the sign-in from Fleetlane’s published OAuth metadata and registers itself.
3. You sign in on the connect screen with your password or an email code, plus your authenticator code when your organisation requires one.
4. The assistant receives a token that is valid only for your workspace’s MCP address and renews it while the connection stays in use.

sequenceDiagram
  actor U as User
  participant A as AI assistant
  participant F as Fleetlane
  U->>A: Add the workspace address with /mcp
  A->>F: Discover sign-in from OAuth metadata
  A->>U: Open the connect screen
  U->>F: Sign in and approve
  F-->>A: Token for this workspace only
  U->>A: Ask about claims
  A->>F: Tool call with the token
  F->>F: Apply the user's roles, scope and field rules
  F-->>A: Results with links back to Fleetlane

Workspaces that require single sign-on. The connect screen signs in by password or email code, so it cannot stand in for your organisation’s enterprise sign-in. If your workspace requires single sign-on, the server refuses a signed-in connection and you connect with a personal API key instead.

Connect with a personal API key

Create a key in your Fleetlane profile under Personal API keys. The full key is shown once. Give it to an assistant or script as Authorization: Bearer <key> (or the X-API-Key header), on your workspace’s MCP address or on the shared key-only address https://mcp.<domain>/mcp, which finds the workspace from the key.

• A personal key acts as you, with your current roles and access, and is logged as you.
• It holds one scope, mcp: it opens the MCP server and its HTTP mirror and nothing else in Fleetlane.
• It expires on the date you choose, and after a year if you choose none. Revoke it in your profile at any time; it also stops working when your access ends.
• Scripts can call the same tools over plain HTTP under /api/v1/mcp/ with the same key.

What the assistant can do

Read. All read tools are read-only and respect the person’s roles, dealership scope and field rules.

Tool	What it does
fleetlane_context	The workspace, the signed-in user and guidance for the assistant
fleetlane_search	Searches the way the app’s search box does: claim and authorisation numbers, VINs, recalls, templates and dealerships, plus what was written in claims and comments where semantic search is on
fleetlane_list_claims	Lists claims and services with search, filters, sorting and pages, optionally only those the user can edit
fleetlane_list_vehicles	Lists vehicles with search, filters, sorting and pages
fleetlane_get_claim	Reads one claim or service in full, with its items and field metadata
fleetlane_get_vehicle	Reads one vehicle in full
fleetlane_get_vehicle_overview	A vehicle’s service plan, outstanding recalls, warranty coverage and claim history
fleetlane_create_options	What the user may create: claim types and dealerships, templates and the code lists a claim takes
fleetlane_get_activity	A claim’s or vehicle’s activity log: changes, lifecycle runs and refused attempts
fleetlane_list_attachments	A record’s attachments and, for a claim, which evidence is still missing
fleetlane_get_discussion	A record’s discussion as plain text
fleetlane_workflow_available_actions	The lifecycle steps the user can run on a claim or vehicle right now, with the input each one expects
fleetlane_workflow_validate_action_payload	Checks a proposed step and its input without running it
fleetlane_workflow_run_status	The outcome of a lifecycle run that was still running

Write. Each write runs the app’s own procedure as the user, so a change the user could not make in the app is refused with the app’s reason. Assistants that support it ask you to confirm every write.

Tool	What it does
fleetlane_raise_claim	Creates a claim of a manual claim type, or from a template
fleetlane_raise_recall_claim	Creates the recall claim for an outstanding recall on a vehicle
fleetlane_log_service	Logs a service from a vehicle’s service plan
fleetlane_update_claim	Edits a claim’s or service’s fields
fleetlane_add_claim_item, fleetlane_update_claim_item, fleetlane_remove_claim_item	Adds, edits or removes a part, labour, sublet or other line; the app prices it
fleetlane_update_vehicle	Edits a vehicle’s fields
fleetlane_post_comment	Posts to a record’s discussion, with mentions
fleetlane_workflow_invoke_action	Runs a lifecycle step, exactly as the app’s action button does

Two guards keep a retry from doing harm. Every edit carries the record’s last-read version and is refused if someone changed the record since, so the assistant reads it again and checks with you instead of overwriting. A lifecycle step carries the status it expects and is refused once the record has moved on, so the same step cannot run twice.

Lifecycle steps run in the lifecycle engine under the published policy, the same as from the app. The step’s checks, rules and calls to your systems decide the outcome, and the run is recorded with the person who started it. The record stays locked while the run is processing. Results link back to the matching record in Fleetlane, so the user can open it and carry on.

What it cannot do

These limits are deliberate:

• It cannot create vehicles or edit a vehicle’s build configuration.
• It cannot upload attachments; files are added in the app.
• It cannot change settings, roles, claim types or lifecycle policies, actions and sequences.
• It cannot skip a lifecycle step or decide its outcome: the published policy does.

There is no AI inside the process. An assistant can press the buttons the person could press; it cannot change what is behind them.