MCP Server (AI Assistants)
Connect AI assistants like Claude directly to the Tire Library catalog using the Model Context Protocol (MCP) — an open standard that lets AI tools query data through natural conversation. Instead of writing code against the REST API, you ask questions: “Find 205/55R16 winter tires from Michelin” or “What rebates are active right now?” — and the assistant calls the catalog for you.
MCP or REST API — which should I use?
| Use case | Recommended access |
|---|---|
| Interactive AI assistants (Claude.ai, Claude Desktop, ChatGPT) used by you or your team | MCP server — browser-based OAuth sign-in, no code to write |
| Production systems, POS/ERP integrations, scheduled jobs, or any headless/unattended agent running on a server | REST API with an API key (tl_live_...) — non-interactive, no re-authentication |
Requirements
- An account with an active API membership subscription (the same tier that unlocks
tl_live_API keys). If you can use the REST API, you can use the MCP server. - An MCP-compatible client: Claude.ai, Claude Desktop, ChatGPT, or any other MCP client.
Connection Details
| Setting | Value |
|---|---|
| Server URL | https://mcp.tireweblibrary.com/sse |
| Authentication | OAuth 2.0 (authorization code + PKCE) — sign in with your Tire Library account email and password |
| Access | Read-only catalog access, tied to your account |
Connect in Claude.ai
- Go to Settings → Connectors → Add custom connector.
- Enter the server URL:
https://mcp.tireweblibrary.com/sse - Click Connect. A browser window opens to our sign-in page.
- Sign in with your Tire Library account credentials.
- Done — Claude can now query the tire catalog on your behalf.
Available Tools
The server exposes 17 read-only tools. Required parameters are marked with *. All tools return JSON; search/list tools return the same paged result shape as the REST catalog endpoints (results.current_page, results.data, etc.). On failure, tools return an error object ({"error": "...", "tool": "...", ...}) instead of throwing.
| Tool | What it does | Parameters | Returns |
|---|---|---|---|
search_tires |
Full filtered catalog search (the most flexible search) | width, aspect_ratio, rim_size, make, speed_rating, load_rating, load_range, ply_rating, utqg, warranty_min, model_name, season, terrain, category, run_flat, three_pmsf, mud_and_snow, studdable, sort, page, per_page — all optional |
Paged tire results |
search_tires_by_size |
Search tires by exact size dimensions | width*, aspect_ratio*, rim_size*; optional make, season, terrain, category, run_flat, three_pmsf, mud_and_snow, studdable, page, per_page |
Paged tire results |
search_by_make |
Search tires from one manufacturer, with optional size/feature filters | make*; optional width, aspect_ratio, rim_size, speed_rating, season, terrain, category, run_flat, three_pmsf, mud_and_snow, studdable, page, per_page |
Paged tire results |
search_by_item_number |
Exact lookup by item/part number | item_number* |
Paged results with the matching tire(s) |
get_tire_by_id |
Full spec detail for one tire SKU | tire_id* |
Tire detail object, or {"found": false} |
compare_tires |
Side-by-side comparison of 2–6 tires | tire_ids* — array of 2–6 unique IDs |
found_count, missing_ids, tires[] |
check_fitment |
Tire sizes that fit a specific vehicle | year*, make*, model*; optional trim |
Fitment data for the vehicle |
list_tire_makes |
All tire manufacturers in the catalog | none | Array of {id, name, image_url, dot_reg_url} |
browse_tire_patterns |
Browse patterns/models across makes and/or by name search | optional make, search, page, per_page |
Paged pattern results |
list_tire_patterns_by_make |
All patterns/models for one manufacturer | make*; optional page, per_page |
Paged pattern results |
get_tire_pattern_by_id |
Pattern detail, including all available sizes | pattern_id* |
Pattern object with tire_sizes[], or {"found": false} |
search_sizes_in_pattern |
Check whether a pattern comes in a given size | pattern_id*; optional width, aspect_ratio, rim_size |
has_matches, match_count, matching_sizes[] |
get_facets |
Available filter values (makes, widths, ratios, rims, ratings, etc.) | optional make |
Facets object from the catalog |
list_rebates |
List manufacturer rebates with optional filters | optional make, status (Active default, Upcoming, Expired, or all), page, per_page |
Paged rebate results |
get_rebate_details |
Rebate detail: form URL, required documents, deadline, reward | rebate_id* |
Rebate detail object, or {"found": false} |
get_rebates_for_tire |
All rebates that apply to one tire SKU | tire_id*; optional is_active (default false) |
rebates[], rebate_count, tire/make/model names |
get_rebates_for_pattern |
All rebates that apply to a pattern (all its sizes) | pattern_id*; optional is_active (default false) |
rebates[], rebate_count, pattern/make names |
Example Prompts
- “List all tire makes and show me Michelin’s winter patterns” —
list_tire_makes,list_tire_patterns_by_make - “What fits a 2021 Ford F-150 XLT?” —
check_fitment - “Are there any active rebates on the Michelin CrossClimate 2?” —
browse_tire_patterns,get_rebates_for_pattern - “Does the Defender come in 225/65R17?” —
search_sizes_in_pattern
Rate Limits
There is no separate MCP quota. Every tool call is forwarded to the REST API, where the standard limit applies: 200 requests per 60-second sliding window under the API tier policy. Most tools translate to a single REST request per call; compare_tires makes one request per tire ID (2–6). If the limit is exceeded, the tool returns an error object with status: 429 — slow down and retry.
Notes
- The data is identical to what the REST API serves — MCP is just a conversation-friendly way to reach it.
- Access is read-only: the assistant can search and view catalog data but cannot modify anything.
- Sign-in sessions renew silently while in use; after ~7 days of inactivity the assistant will ask you to sign in again (see the callout at the top of this page).