Skip to main content
Glama
Compass DaaS

Compass DaaS

Official

Compass DaaS MCP

Smithery badge

MCP server for Compass DaaS. It exposes three read-only restaurant tools (search, enrich, decide fit) over the Model Context Protocol and calls the Compass REST API with your API key. Coverage is the 50 US states and Washington, DC.

Compass is not an allergen or safety service.

This package replaces @compass-food/mcp, which is no longer updated. Change @compass-food/mcp to @compassfoodtech/mcp in your MCP client config.

Quick Start

npx -y @compassfoodtech/mcp

Set COMPASS_API_KEY before starting the server:

export COMPASS_API_KEY="<paste the key from signup>"

For staging or local testing, set:

export COMPASS_BASE_URL=https://daas-api-veganmapai-1a8b7.a.run.app

Related MCP server: Maître d'MCP

Install

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "compass": {
      "command": "npx",
      "args": ["-y", "@compassfoodtech/mcp"],
      "env": {
        "COMPASS_API_KEY": "<paste the key from signup>"
      }
    }
  }
}

Cursor

Add to Cursor MCP settings (~/.cursor/mcp.json globally, or .cursor/mcp.json in a workspace):

{
  "mcpServers": {
    "compass": {
      "command": "npx",
      "args": ["-y", "@compassfoodtech/mcp"],
      "env": {
        "COMPASS_API_KEY": "<paste the key from signup>"
      }
    }
  }
}

Codex CLI

Edit ~/.codex/config.toml:

[mcp_servers.compass]
command = "npx"
args = ["-y", "@compassfoodtech/mcp"]

[mcp_servers.compass.env]
COMPASS_API_KEY = "<paste the key from signup>"

Get an API key

Sign up — a working sandbox key is saved immediately, with 250 credits for 72 hours. Confirming email raises the shared free allowance to 1,000 credits per month.

Tools

Wraps POST /v1/search. The mode tool argument is sent to the REST API as the X-Compass-Mode header. Pass a dish in dish and the destination as place or location, not both. Each result carries vegan_dining (up to five relevant vegan dishes and menu_collected_at). A dish query with no restaurant that has the dish returns result_status: "no_match" and, when one exists within 50 km, nearest_match. Set include_cuisine_matches to also list same-cuisine restaurants without the dish, marked cuisine_only.

{
  "query": "ramen",
  "dish": "ramen",
  "place": { "country_code": "US", "region": "NY", "city": "Brooklyn" },
  "user_profile": { "diet": "strict_vegan" },
  "limit": 5
}
{
  "query": "vegan ramen in Brooklyn",
  "user_profile": {
    "diet": "strict_vegan",
    "exclude_cross_contamination": true
  },
  "location": {
    "lat": 40.6782,
    "lng": -73.9442,
    "radius_m": 5000
  },
  "limit": 10,
  "mode": "rich"
}

compass_enrich_restaurant

Wraps POST /v1/enrich/restaurant. Use compass_id or google_place_id for an exact lookup (one of them, not both), name plus address, or name plus place. matched is true only when identity_status is resolved; otherwise match_candidates lists possible restaurants.

{
  "name": "Buddha Bodai",
  "address": "5 Mott St, New York, NY"
}
{
  "name": "Blossom",
  "place": { "country_code": "US", "region": "NY", "city": "New York" },
  "dish": "seitan"
}

compass_decide_fit

Wraps POST /v1/decision/restaurant-fit. The mode tool argument is sent to the REST API as the X-Compass-Mode header. Send exactly one of compass_id or google_place_id. vegan_options is an alias of strict_vegan; set fully_vegan_only to accept only fully vegan restaurants. vegetarian is judged from vegan evidence; other accepted diets return unknown with DIET_NOT_SUPPORTED. exclude_cross_contamination defaults to false. Send true when unknown shared-prep evidence should return unknown. When a google_place_id maps to no restaurant or to several, the response has resolution and coverage instead of a decision.

{
  "compass_id": "rest_xyz789",
  "user_profile": {
    "diet": "strict_vegan"
  },
  "mode": "rich"
}

API Key Resolution

The server checks:

  1. COMPASS_API_KEY

  2. ~/.compass/config.json

Config file:

{
  "api_key": "<paste the key from signup>",
  "base_url": "https://api.compassfoodtechnologies.com"
}

Privacy

This package sends tool calls only to the configured Compass API base URL. It does not send secondary usage data.

Available Tools

3 tools
compass_decide_fitDecide restaurant fitA
Read-onlyIdempotent

Conservative fit decision for a restaurant against a user dietary profile. Returns "fit", "not_fit", or "unknown" with reason codes, evidence, and user wording. Returns "unknown" rather than overclaiming.

NameRequiredDescriptionDefault
modeNoResponse detail mode; rich includes fuller evidence and reasoningrich
compass_idYesCompass restaurant ID, obtained from compass_search or compass_enrich_restaurant
user_profileYesRequired dietary profile for the restaurant fit decision

Output Schema

NameRequiredDescription
decisionYesConservative decision for the submitted profile
evidenceYesEvidence supporting the decision
compass_idYesStable Compass restaurant ID
confidenceYesDecision confidence
risk_flagsNoRisk flags that should be surfaced to the user
reason_codesYesCompass public reason codes explaining the decision
source_freshnessNoAge and refresh guidance for the evidence behind a result
last_evaluated_atNoWhen Compass last evaluated this restaurant for decision freshness
recommended_user_textNoConservative user-facing wording
verification_requiredNoTrue when the user should verify with the restaurant before relying on the result
compass_enrich_restaurantEnrich restaurantA
Read-onlyIdempotent

Match a restaurant by name and address, then return Compass enrichment data including VeganScore, vegan dietary profile, and evidence. Does not return certification/free-from facts. Returns "matched: false" with candidates if confidence is below threshold.

NameRequiredDescriptionDefault
nameNoRestaurant name
addressNoStreet address (improves match)
compass_idNoCompass restaurant ID for direct lookup
google_place_idNoGoogle Place ID (highest match confidence)

Output Schema

NameRequiredDescription
matchedYesWhether Compass confidently matched the submitted restaurant
evidenceNoEvidence for the enrichment result
candidatesNoCandidate matches returned when confidence is below threshold
compass_idNoStable Compass restaurant ID for the matched restaurant, when matched
confidenceNoMatch confidence when available
restaurantNoCompass restaurant result
vegan_scoreNoDeterministic Compass VeganScore summary
dietary_profileNoPublic restaurant dietary facts supported by launch evidence. Compass does not expose allergen-safe or certification claims.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

  1. 3 tool updatesv0.1.0
    • First observedcompass_decide_fit
    • First observedcompass_enrich_restaurant
    • First observedcompass_search

TDQS

A4.4/5.0

Scored across 3 tools

Disambiguation5/5

Each tool has a distinct purpose: decide_fit makes a dietary fit determination, enrich_restaurant returns enrichment data for a specific restaurant, and search allows natural-language queries. There is no overlap or ambiguity.

Naming Consistency5/5

All tools follow the consistent pattern 'compass_verb_noun' (decide_fit, enrich_restaurant, search). The naming is predictable and uniform.

Tool Count5/5

With only 3 tools, the set is appropriately scoped for the narrow domain of restaurant dietary matching. Each tool is essential and there is no bloat.

Completeness5/5

The tools cover search, enrichment, and fit decision, which are the core operations for the domain. No obvious gaps exist given the stated purpose of providing dietary data service.

Maintenance

ActivityNo data
ResponsivenessOutdated data

Related MCP Connectors

Related MCP Servers

  • A
    license
    B
    quality
    D
    maintenance
    MCP Compass is a discovery and recommendation service that assists AI assistants in finding and understanding Model Context Protocol servers through natural language queries.
    1
    200 npm
    248
    MIT
  • F
    license
    A
    quality
    D
    maintenance
    An MCP server for restaurant discovery and booking across Resy and OpenTable via natural language. It integrates Google Places data with dietary preferences, visit history, and weather awareness to provide personalized dining recommendations and group reservation management.
    23
    -
  • F
    license
    Not graded
    quality
    Not graded
    maintenance
    An AI-native restaurant discovery service that enables searching and receiving natural language recommendations for over 2,200 restaurants across 15+ US cities. It provides tools for accessing detailed restaurant info, curated lists, and cuisine-specific searches through the Model Context Protocol.
    -
  • A
    license
    Not graded
    quality
    D
    maintenance
    Personalized restaurant recommendations, table bookings, and delivery via MCP, CLI, or API, learning user taste and acting proactively.
    MIT