AetioReferral MCP Server
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| LLM_PROVIDER | No | The pydantic-ai provider to use for the optional generate path, such as Anthropic, OpenAI, Google Gemini, OpenRouter or Ollama. Without an LLM configured, only the render tool is available. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| render_referral_letterA | Render a referral letter from a referring doctor to a specialist as a PDF. Write the letter content like a real referral letter: every section is plain text in the letter language, and empty sections are left out. Leave patient, sender and recipient fields empty to have them generated; set patient to null for a letter without patient details. Salutation, opening sentence, closing, signature and stamp are added automatically. Returns the PDF as an embedded resource and the completed letter (with generated fields) plus the seed as structured content. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 1 tool
With a single tool there is no possibility of selection confusion; the one tool's purpose (rendering a referral letter PDF) is unambiguous and clearly stated.
The lone name render_referral_letter follows a clean verb_noun snake_case convention, so there is no inconsistency to penalize.
A one-tool server is borderline thin; while the tool is non-trivial and covers a full render pipeline, the surface offers no companion operations for a domain that plausibly needs more than a single call.
The core action of producing a referral letter PDF is covered, but there is no way to list, retrieve, validate, or re-render existing letters, leaving agents without lifecycle operations around the generated output.