chore: version packages

[github-actions]

This PR was opened by the Changesets release GitHub action. When you're ready to do a release, you can merge this and publish to npm yourself or setup this action to publish automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to v2/main, this PR will be updated.

Releases

@modelcontextprotocol/server-everything@1.0.1

Patch Changes

• #5014 8ea5e05 Thanks @cliffhall! - Fix the async task tools: trigger-elicitation-request-async now stops polling before the 10-minute TTL it requests for the client's task runs out, instead of polling past it and failing on a task the client has expired (#4986); and trigger-sampling-request-async reports the status message of a client task that is already finished when it is created, instead of "No message" (#4987). trigger-elicitation-request-async had the same flaw and gets the same fix.

• #5005 63dc92f Thanks @cliffhall! - Label dynamic blob resources application/octet-stream, matching their resource template, instead of text/plain. This applies to resources/read, the get-resource-reference and get-resource-links tools, and the resource-prompt prompt; blob resource links are now described as "binary blob resource".

• #5034 6bed4c5 Thanks @cliffhall! - Server instructions no longer tell every client to use tools that are registered only for clients declaring a capability; they list those tools and the capability each needs (#4792, ported from #4835). trigger-elicitation-request and trigger-elicitation-request-async are now registered only for clients that support form-mode elicitation, so a client that declared only URL-mode elicitation no longer sees tools it cannot answer (#4985).

• #5053 c63255a Thanks @cliffhall! - Stop shipping the compiled test suite and Vitest config in the published package: the build now excludes __tests__/, *.test.ts, *.spec.ts and vitest.config.ts, as the other TypeScript servers' builds do.

• #5032 d4fb2f4 Thanks @cliffhall! - Session resources are now tracked per server, so two sessions that create a session resource with the same name (for example via gzip-file-as-resource) no longer evict each other's resource (#4808).

• #5020 f7af617 Thanks @cliffhall! - The HTTP+SSE transport now answers requests for sessions it cannot serve instead of leaving them hanging or failing with a 500: POST /message for an unknown session returns 404 (and 400 with no sessionId), and GET /sse?sessionId=… returns 409 for a session that already has its stream and 404 for an unknown one. Each carries a JSON-RPC error body.

• #5006 dbd1b2e Thanks @cliffhall! - Streamable HTTP: answer an unknown or ended session ID with 404 Not Found (carrying the request's id on a POST) instead of 400; replay only the resumed stream's events after a Last-Event-ID, and refuse an unknown one; close every open session on shutdown.

• #4970 c449b5e Thanks @cliffhall! - Internal refactor for in-process testing, with no change in behavior: the launcher only starts a transport when run as the binary, and the stdio, SSE and Streamable HTTP transport modules export their start-up (startStdioServer, startSseServer, startStreamableHttpServer, and createApp() for the HTTP transports) instead of starting a server when imported.

• #4936 feb251d Thanks @cliffhall! - The HTTP+SSE transport no longer prints "Server is running" and stays up when its port is already in use: it reports the port and exits non-zero, as Streamable HTTP does. Streamable HTTP no longer prints its listening line before that error.

@modelcontextprotocol/server-filesystem@1.0.1

Patch Changes

• #4970 3a63f52 Thanks @cliffhall! - Internal: the server is now built by a createServer() factory in server.ts, and index.ts starts it over stdio only when run as the bin, with each server instance holding its own allowed directories. No change to the tools, their results, or how the allowed directories are resolved from arguments and Roots.

• #5011 d8ebdbe Thanks @cliffhall! - edit_file keeps a file's line endings: a CRLF file is written back with CRLF instead of being converted to LF. When oldText has no exact match and the whitespace-tolerant matcher is used, each replacement line now takes the indentation of the file line it replaces, shifted by its indentation relative to the matching oldText line, instead of only the first line being reindented.

• #5022 7dc802f Thanks @cliffhall! - write_file and edit_file now overwrite an existing file in place instead of renaming a temp file over it, so the file keeps its inode, creation time (birthtime), hard links and permission bits (#4512), and an overwrite no longer fails with EPERM on Windows when another process holds the file open (#3199). A symlink swapped in after the path was validated is still refused rather than written through, now by an O_NOFOLLOW open (on POSIX) and a check that the opened file is the one validated (on every platform). The overwrite is no longer crash-atomic. A read-only file is now refused (EACCES, or EPERM on Windows) instead of being replaced, since writing in place opens the file itself for writing.

• #5025 2309e08 Thanks @cliffhall! - Tool calls now wait for the client's initial roots before checking paths, so a call sent right after connecting is no longer refused with "Access denied" (#3204). A server started with no directories, connected to a client without Roots support, now fails visibly: it logs the reason, closes the connection and exits with status 1, instead of staying up and refusing every call (#4992).

• #5030 5c04d74 Thanks @cliffhall! - Every tool's input and output schema in tools/list now declares "$schema": "https://json-schema.org/draft/2020-12/schema" instead of draft-07, so clients that validate tool schemas strictly against JSON Schema 2020-12 no longer reject every tool. The schemas are otherwise unchanged.

• #5010 fd4dff7 Thanks @cliffhall! - On Windows, a UNC share root given as an allowed directory (\\server\share or \\server\share\) now admits the files and subdirectories under it. Before, only the share root itself was accessible and every path below it was refused. Sibling shares such as \\server\share-evil are still refused.

• #5009 f621c75 Thanks @cliffhall! - Paths that differ from the names on disk only in Unicode form now resolve (#1970). An NFC spelling of an allowed directory whose name is stored NFD (macOS "Capture d’écran", Japanese dakuten) is no longer refused as outside the allowed directories, and a request that spells a U+202F or U+00A0 in a name as a plain space (the macOS screenshot "Screenshot … at 2.40.40 PM.png") finds the file instead of failing with ENOENT. The exact spelling still wins when it exists, and a name matching two entries is refused as ambiguous.

@modelcontextprotocol/server-memory@1.0.1

Patch Changes

• #5008 bb8aa6f Thanks @cliffhall! - create_entities now reports the entities it skipped because their name already exists (or repeats earlier in the same call): the structured result lists them in a new optional skipped array, a second text item names them and points to add_observations, and the tool description says so. Skipped entities are still not created and their observations are still not added (#4887).

• #4970 a8127ba Thanks @cliffhall! - Build the server with an exported createServer() factory and start stdio only when the package is run as a program, so importing the module no longer starts a server. Behavior over stdio is unchanged.

• #5037 7be08e9 Thanks @cliffhall! - Two server processes sharing one memory file no longer silently discard each other's writes: every write tool now holds an exclusive lock file (<memory file>.lock) while it reads, changes and saves the graph, and a lock left by a crashed server is recovered automatically (#4797).

• #4937 478db6a Thanks @cliffhall! - Internal: ensureMemoryFilePath accepts the directory its default files live in, so the server's tests no longer write into the package directory. No change to how the server chooses or migrates its memory file.

• #5035 2b1305f Thanks @cliffhall! - Lines in the memory file that the server cannot read (malformed JSON, an entity or relation that fails validation, an unknown record type) are no longer deleted by the next write. They are still left out of the graph, and are now written back unchanged after the graph's own lines.

• #5053 c63255a Thanks @cliffhall! - Stop shipping the compiled test helper dist/__tests__/helpers.js in the published package: the build now excludes everything under __tests__/, not just *.test.ts.

• #5013 819ae90 Thanks @cliffhall! - Saving the knowledge graph keeps the memory file's permission bits (an operator's 0600 no longer comes back 0644), and a write to a read-only memory file now fails with EACCES instead of silently replacing it (#4827).

@modelcontextprotocol/server-sequential-thinking@1.0.1

Patch Changes

• #5015 8a4139c Thanks @cliffhall! - Advertise the sequentialthinking tool as readOnlyHint: false and idempotentHint: false. Every call appends to the server's thought history, and a call with both branchFromThought and branchId also appends to that branch, so the tool was never read-only or idempotent. destructiveHint and openWorldHint stay false. (#4721)

• #4970 86806cc Thanks @cliffhall! - Build the server with an exported createServer() factory and attach stdio only when index.js runs as the binary, so the server can be tested in-process. No change to the tool, its schemas or its results.

• #5031 bc52d0e Thanks @cliffhall! - The thought log on stderr no longer prints "undefined" in its headers: a revision with no revisesThought reads "Revision N/M", and a branch with no branchId reads "(from thought N)" with no ID. The box border is sized from the header's visible text, so colour escape codes no longer make it wider than its text.

• #5053 c63255a Thanks @cliffhall! - Stop shipping the compiled test helper dist/__tests__/helpers.js in the published package: the build now excludes everything under __tests__/, not just *.test.ts.

• #5036 b5c5ccc Thanks @cliffhall! - A branchId that names an Object.prototype key, such as "constructor" or "__proto__", now creates and lists its branch like any other id instead of failing the call. A call that fails no longer adds its thought to the history first, so thoughtHistoryLength counts only thoughts that were accepted.