Skip to content
IsoKron

Directory review rules for MCP tools

Both directories review the tool list itself: tool names, titles and hints, descriptions that say what a tool does and when to use it, separate read and write tools, and no inputs for secrets, precise location or conversation history.

A submission to the Claude connector directory or the ChatGPT app directory is reviewed against a written checklist, and much of it can be read straight off the answer to tools/list. These checks apply the tool-level rules of both documents as they read on the date shown on each page.

A pass here does not mean approval: reviewers also test the server, its authentication and its listing. What it does show is that the tool-level rules these checks cover are met.

The 24 checks

  • Reads and writes are separate tools: A single tool that accepts both safe methods (GET, HEAD, OPTIONS) and unsafe ones (POST, PUT, PATCH, DELETE) is rejected by the Anthropic directory, because no hint can be true for it; split it into a read tool and write tools.
  • Custom query tools name or link the target API: If a tool accepts free-form endpoint paths, query strings or request bodies, its description must link to or name the target API; the Anthropic checklist gives "Makes a request to the API" as a description that fails.
  • No inputs for memory, chat history or conversation data: Tools must not ask for Claude’s memory, chat history, conversation summaries or user files, so arguments named for conversation history, transcripts or memory block Anthropic review; take a short task-specific field instead.
  • Not an unsupported use case (money transfer, AI media generation): The Anthropic directory does not accept connectors that transfer money, cryptocurrency or other financial assets, or that generate images, video or audio through AI models; diagrams, charts and UI mockups are allowed.
  • No conversation history or "just in case" context inputs: OpenAI’s guidelines forbid requesting or reconstructing full conversation histories or collecting broad context "just in case", so history and transcript arguments block review and generic context fields are flagged.
  • No credential, payment card or government ID inputs: OpenAI lists credentials (passwords, API keys, MFA/OTP codes), payment card data and government identifiers as restricted data a plugin must not collect, so a tool argument for any of them blocks review; take credentials through OAuth.
  • No precise location inputs: OpenAI says to rely on the coarse location the system shares and not to request precise user location, so latitude, longitude or GPS arguments block review; addresses are flagged unless they are clearly task data.
  • No generic executor or operation dispatcher: OpenAI does not allow discovery, operation selection or schema fetching with a generic executor that enables operations not individually exposed for review; expose each operation as its own tool.
  • Every tool has a title: Every tool needs a short human-readable title, such as "Create issue", for the Claude connector directory; either the top-level title field or annotations.title counts.
  • Every tool carries the applicable hint: The Anthropic connector checklist requires every tool to carry the applicable hint: readOnlyHint: true on read-only tools and destructiveHint: true on tools that modify or delete data; Claude uses them to decide which calls run without confirmation.
  • destructiveHint: false on a write: the platforms word this differently: OpenAI says destructiveHint: false is right for additive writes with no destructive or irreversible effect; Anthropic asks for destructiveHint: true on tools that modify or delete data and does not say whether a pure create counts, so false is defensible only for writes that purely add.
  • Tool names of 64 characters or fewer: The Anthropic connector checklist requires tool names of 64 characters or fewer, and Claude validates every name against ^[a-zA-Z0-9_-]{1,64}$, so a longer name, or one with a dot or space, fails.
  • Description says what the tool does and when to use it: Each tool description should state precisely what the tool does and when to invoke it; a missing description blocks Anthropic review, and one under about 40 characters rarely covers both.
  • readOnlyHint, destructiveHint and openWorldHint are explicit booleans: OpenAI requires readOnlyHint, destructiveHint and openWorldHint as explicit true or false booleans in every tool’s annotations, even where the MCP specification has a default; the submission form will not continue until they are set.
  • Every tool has a non-empty description: OpenAI requires every tool to have a non-empty description that explains its purpose and behavior, when to use it, and any limitations or side effects.
  • Tool names are unique within the server (ChatGPT): OpenAI’s plugin guidelines state that tool names must be unique within your MCP server, so a duplicate name blocks a ChatGPT app submission even though the MCP specification only says SHOULD.
  • Tool names without promotional or comparative words: OpenAI’s guidelines say tool names must avoid misleading, overly promotional or comparative language (for example pick_me, best, official); name the action plainly instead, such as get_order_status.
  • Tool names that say what the tool does: OpenAI’s guidelines ask you to avoid internal jargon, unexplained acronyms and opaque identifiers in tool names; use a plain verb and object such as list_invoices.
  • Write tools say whether a retry repeats the effect: OpenAI asks that tools be safe to retry where possible, or say when a retry repeats the effect; mark naturally repeatable writes with idempotentHint: true, and say so in the description of writes that are not.
  • Tools that start jobs or queue work are not read-only: A tool that starts a job, queues work, schedules something or persists an artifact changes external state, so OpenAI asks for readOnlyHint: false even if the tool returns only a job ID.
  • Read-only tools set destructiveHint: false: A tool with readOnlyHint: true should set destructiveHint: false: the specification ignores destructiveHint on read-only tools, but the pair contradicts itself and OpenAI asks for false.
  • Deleting tools are not marked non-destructive: Tools that delete, overwrite, cancel, revoke or reset must set destructiveHint: true; with false, clients run them without a confirmation prompt, and both directories treat that as incorrect annotations.
  • Tools named as writes are not marked read-only: A tool whose name reads as a create, update or delete must not set readOnlyHint: true, or clients run it without asking; set readOnlyHint: false and an explicit destructiveHint unless the tool only previews.
  • No steering of when other tools are used: A tool description must not steer the model toward or away from other servers’ tools ("always call this tool first", "never use other tools"); OpenAI and Anthropic both reject it, while routing between your own tools is fine.

Other topics

Sources last verified on 2026-11-18. How findings become grades.

Check your tool list: paste your tools/list, nothing is sent anywhere