LLMIntel
OverviewCatalogCompareMCPPricingChangelogQuickstartDocs
Dashboard
LLMIntel

A verified, normalized feed of AI model lifecycle events across every major provider, so a deprecation shows up on your roadmap instead of in your error logs. Plus runtime telemetry for what those models actually cost you.

Questions? support@llmintel.ai

Product
  • Pricing
  • Compare models
  • Dashboard
  • Changelog
  • Contact support
Developers
  • Quickstart
  • Model catalog
  • MCP server
  • API reference
  • OpenAPI spec
  • @llmintel/telemetry
Free & open data

The model catalog is a normalized, machine-readable source of truth for AI model lifecycle events. It is free via the public API, and every record links to the primary provider source it came from.

© 2026 Alex Tsimbalistov. All rights reserved.
LLMIntel
OverviewCatalogCompareMCPPricingChangelogQuickstartDocs
Dashboard

Stop your coding agent shipping a dead model id

Every model is trained on a snapshot of the world, so agents confidently write model ids that stopped answering months ago. This MCP server gives yours a live lookup against LLMIntel's verified lifecycle catalog, covering OpenAI, Anthropic, Azure AI Foundry, AWS Bedrock, Google, and Cohere.

No API key. No account. The catalog is public.

Install

Add it to any MCP host. It runs straight from npm, so there is nothing to install first.

Cursor

Add to .cursor/mcp.json in your project (or the global config):

{
  "mcpServers": {
    "llmintel": {
      "command": "npx",
      "args": ["-y", "@llmintel/mcp"]
    }
  }
}

Claude Code

claude mcp add llmintel -- npx -y @llmintel/mcp

Claude Desktop

Same block as Cursor, in claude_desktop_config.json.

Hosted endpoint

Hosts that take a URL need no Node and no package. The same five tools are served over Streamable HTTP:

{
  "mcpServers": {
    "llmintel": {
      "url": "https://llmintel.ai/v1/mcp"
    }
  }
}

What your agent can ask

ToolAnswers
check_modelIs this model id safe to use? Returns lifecycle state, the retirement deadline in days, the recommended replacement, and the provider source link.
list_retiring_modelsWhat breaks in the next N days? Models already past their retirement date are listed separately, first.
suggest_replacementWhat do I migrate to? The provider's own recommendation where one exists, otherwise same-provider active models clearly labelled as candidates.
search_modelsWhat can I use from this provider right now? Filter by provider and lifecycle state.
recent_lifecycle_changesWhat was deprecated recently? The cross-provider change feed.

Example

Ask your agent to audit the model ids in a directory. For each one it calls check_model and gets back:

ACT NOW — this model is retiring imminently.

"anthropic.claude-3-5-sonnet-20240620-v1:0" resolves to the
tracked model bedrock/anthropic.claude-3-5-sonnet-20240620-v1-0.

Provider: bedrock
Lifecycle state: retiring — retirement scheduled
Deprecated: 2026-01-30 (179 days ago)
Retirement: 2026-07-30 (in 2 days)

The provider has not named a replacement.
Use suggest_replacement for options.
Pricing/limits: $3/1M in · $15/1M out

Source: docs.aws.amazon.com/bedrock/latest/userguide
        /model-lifecycle.html
Provider's own term: "end of life"

Deadlines are always expressed in days, because a language model cannot reliably judge whether 2026-07-30 is soon.

How it avoids lying to your agent

  • If the catalog is unreachable, the tool returns an MCP error rather than an answer. An agent that read a network failure as “no deprecation found” would happily ship a retired id.
  • A model we don't track returns “verify with the provider”, never “OK”.
  • Replacements the provider published are labelled as such; our own fallback suggestions are marked as candidates to evaluate.
  • Anything past its retirement date is broken now, so it never sits buried in a “retiring soon” list.

Configuration

Both environment variables are optional. Anonymous callers get 30 requests/minute per IP, which is ample for interactive use.

  • LLMINTEL_API_KEY raises the rate-limit budget. You do not need one.
  • LLMINTEL_BASE_URL points at a self-hosted or staging catalog.
Browse the catalog →REST API reference
LLMIntel

A verified, normalized feed of AI model lifecycle events across every major provider, so a deprecation shows up on your roadmap instead of in your error logs. Plus runtime telemetry for what those models actually cost you.

Questions? support@llmintel.ai

Product
  • Pricing
  • Compare models
  • Dashboard
  • Changelog
  • Contact support
Developers
  • Quickstart
  • Model catalog
  • MCP server
  • API reference
  • OpenAPI spec
  • @llmintel/telemetry
Free & open data

The model catalog is a normalized, machine-readable source of truth for AI model lifecycle events. It is free via the public API, and every record links to the primary provider source it came from.

© 2026 Alex Tsimbalistov. All rights reserved.