Bragproof

← Reports

@myspec/mcp-server

Starts with no setup1.1.2npm

Installed, started with no arguments, no environment and no input, and listed its tools. Run 2026-11-16.

Needs to start: nothing
Returns a result when called with no setup: 0 of 2 tools tried
Contacts while installing and starting: nothing

45tools listed
0hosts contacted
0files written
0.4 sto start
186dependencies

What it contacted

Nothing besides the npm registry it was installed from. Watched from outside its sandbox while it was installed and started with no input. Its tools may still contact things when called.

Starting it

npx -y @myspec/mcp-server@1.1.2

Installing a package runs its code. Use a sandbox for anything you have not read.

Calling its tools

2 tools that take no input and look read-only were called once with nothing. 0 returned a result.

list_projectsNo usable answer tool reported an error: list_projects failed: Could not discover endpoints from https://app.myspec.dev/api/v1/config (Not authenticated. Set MYSPEC_API_TOKEN to
list_stream_tokensNo usable answer tool reported an error: list_stream_tokens failed: Could not discover endpoints from https://app.myspec.dev/api/v1/config (Not authenticated. Set MYSPEC_API_TOK

During the calls it contacted nothing it had not already contacted.

Only tools whose names read like looking something up, with no required input, are tried, five at most. Tools that need input, and anything that sounds like it changes or sends something, are not called.

What its tools tell an agent

In the tool descriptions, nothing that talks about the user, other tools or private files, or uses hidden characters.

Found by simple pattern rules in each tool's description (up to 3,000 characters), quoted as written. Parameter descriptions are not scanned. A clean result is not a safety review.

Tools

list_projects query, limit, offset
List or search MySpec projects accessible to the authenticated user.
get_project project_id
Get details for a project including all active spec sessions, files (id + metadata), and attachments (id + metadata).
get_spec_file file_id, include_download_url
Get spec file metadata. With `include_download_url=N`, also returns a signed, time-limited download URL for revision N (latest or historical). The URL is self-authenticating, so no credentials are returned.
read_spec_file file_id, revision, offset, limit
Return a window of a spec file's UTF-8 content, paginated by line (offset/limit, default and cap 2000 lines, 1 MiB per response). Large files are read across multiple calls using `next_offset`. Bytes are cached on disk under the configured cache root (default ~/.myspec) at `project/<projectId>/file/<fileId>/rev/<rev>/<basename>` and reused on subsequent reads of the same revision, so paging through a file downloads it once. Files above 8 MiB and binary content are refused — use download_spec_file to save those to a local file instead.
download_spec_file file_id, revision, destination_path, overwrite
Download a spec file's raw bytes to a local file. By default it lands under `.specs/` in the MCP server's working directory, mirroring the file's remote path — pass `destination_path` only to save it somewhere else. Unlike read_spec_file, this handles binary files and files up to 50 MiB, and the response returns only metadata (saved path, size, revision) — not the content. Use it to save a large or binary file locally, then search or read it in chunks with other tools.
40 more tools
list_spec_file project_id, path, trashed, limit, offset
List a project's spec files with id, path, and metadata (revision count, file size of the latest revision, file type, owning session, last update). Filter to a directory with `path` (exact file or everything under it; pass "", "/", or omit for all). Set `trashed=true` to list files in the Trash Bin instead of live files. Ordered by `file_path` ascending whenever `path` or `trashed` is given, most-recently-created first otherwise. Returns `total` and, when more records remain, `next_offset`.
get_attachment attachment_id, include_download_url
Get attachment metadata. With `include_download_url>0`, also returns a signed `download_url`.
upload_attachment project_id, file_path, file_name, mime_type, override
Upload a local file as a new attachment on a project. Reads `file_path` from the host running the MCP server, computes its SHA256, then performs the two-phase platform upload. Max 10 MB. Supported content: PDF, DOCX, XLSX, images (PNG, JPEG, GIF, WEBP), and any UTF-8 text file (XML, JSON, YAML, HTML, SVG, Markdown, CSV, source code, ...). Other binary files are rejected by the platform.
create_project name, description
Create a new project. Returns the created project id and metadata.
update_project project_id, name, description, project_type
Update a project's metadata (name, description, project_type). Returns the updated metadata.
archive_project project_id
Archive a project (sets status to "archived"). Idempotent: archiving an already-archived project succeeds.
unarchive_project project_id
Unarchive a project (sets status to "active"). Idempotent: unarchiving an already-active project succeeds.
delete_project project_id
Soft-delete a project. Only a project that is already archived can be deleted. If the project is still active, this tool refuses and you must NOT archive it on the user's behalf to make the deletion go through — instead, reconfirm with the user whether deleting the project is truly intended.
upload_spec_file project_id, file_path, content, local_file_path, file_type
Upload a new, session-less spec file into a project. Provide the file body inline via `content` or by pointing at a local file with `local_file_path` — exactly one is required. Prefer `local_file_path` when the body is already an existing local file (the server reads it directly, avoiding sending large content through the model); use `content` for text generated in-memory. Rejects if a file already exists at the same path (use update_spec_file to overwrite).
update_spec_file file_id, project_id, file_path, content, local_file_path, expected_version
Overwrite an existing spec file by saving a new revision. Provide the new body inline via `content` or by pointing at a local file with `local_file_path` — exactly one is required. Prefer `local_file_path` when the body is already an existing local file (the server reads it directly); use `content` for text generated in-memory. Reference the file by file_id or by (project_id + file_path). Pass `expected_version` from the get_spec_file you based the new body on so a concurrent edit by someone else is reported instead of silently overwritten.
move_spec_file_to_trash file_id
Move a spec file to the Trash Bin. This is recoverable — the file is hidden from the default file list and AI-agent tools but kept restorable indefinitely via `restore_spec_file_from_trash`. It is NOT a permanent delete.
move_spec_folder project_id, from_path, to_path
Move or rename a folder under specs/ together with everything in it. The folder can be a bundle (it holds spec files directly) or a group folder holding bundles. The files of a bundle always move together; single files cannot be moved out of a bundle. File ids, contents and revisions stay the same; only paths change, so address the files by their new paths afterwards. Refused: OpenSpec folders, specs/memory, a target inside the folder itself, a target that already exists (Trash included) or sits inside a bundle, a folder inside a bundle (move the bundle instead), a folder whose bundle is still generating, and a move that would nest a bundle more than 3 levels below specs/.
restore_spec_file_from_trash file_id
Restore a spec file from the Trash Bin back to the active file list, reversing `move_spec_file_to_trash`. The file becomes visible again to the default file list and AI-agent tools.
list_attachments project_id, limit, offset
List a project's attachments (metadata only, always including file_size_bytes so a caller can check the size before read_attachment). Returns `total` and, when more records remain, `next_offset`.
read_attachment attachment_id, offset, limit
Return an attachment's content inline. UTF-8 text is paginated by line (offset/limit, 1 MiB per response). Images (png, jpeg, gif, webp) are returned whole as an MCP image block — never paginated — up to the 10 MiB attachment limit; a 10 MiB image is roughly 13.4 MiB of base64, so check file_size_bytes (from get_attachment or list_attachments) before reading a large one. Other binary types (pdf, docx, xlsx, …) cannot be read here; view them in the MySpec webapp.
list_spec_sessions project_id, archive_filter, status, query, limit, offset
List a project's spec sessions (metadata plus session_summary and context_size_bytes — never the context itself or the chat transcript). Filter with archive_filter ('active'/'archived'/'all'), status, and query (searches the session summary). Capped at 20 per page; returns `total` and `next_offset`. context_size_bytes here is measured by the platform and may read a few percent above get_spec_session's, which is the figure checked against the include_context ceiling.
get_spec_session session_id, include_context
Get one spec session's metadata: status, counters, timestamps, archived_at (delete eligibility), session_summary, plus context_size_bytes and context_keys describing the context without returning it. Set include_context=true for the full context (1 MiB ceiling). The chat transcript is never returned.
rename_spec_session session_id, session_summary
Rename a spec session by updating its context.sessionSummary — the session's human-readable title (there is no first-class name field). Permitted on both active and archived sessions. All other context keys are preserved untouched.
archive_spec_session session_id
Archive a spec session. Archiving is the confirmation step required before delete_spec_session, and is reversible via unarchive_spec_session. Idempotent: archiving an already-archived session succeeds and keeps the original archived_at.
unarchive_spec_session session_id
Unarchive a spec session, returning it to the active list.
delete_spec_session session_id
Soft-delete a spec session. Only an ARCHIVED session can be deleted — archive it first with archive_spec_session. This tool never archives on your behalf: archiving is the user's confirmation step, and collapsing the two-step destructive flow would remove the only checkpoint the platform provides. Check archived_at via get_spec_session to see whether delete is permitted.
list_artifacts project_id, limit, offset
List a project's live artifacts (multi-file bundles such as HTML mockups, each versioned as one snapshot) with id, slug, entry path and revision_count. Returns `total` and, when more remain, `next_offset`. Use get_artifact for the file manifest and read_artifact_file for content.
get_artifact artifact_id, revision
Get an artifact's metadata plus the file manifest (path, content type, size, checksum of every file) of its head revision, or of `revision` when given. `revision_count` is the head number AND the value to pass as expected_version when writing; it only ever grows because rollback appends a revision rather than discarding history.
read_artifact_file artifact_id, path, revision, offset, limit
Return one artifact file's UTF-8 content, paginated by line (offset/limit, default and cap 2000 lines, 1 MiB per response) with `total_lines`, `truncated` and `next_offset`. Reads the head revision unless `revision` is given. Binary files (images, fonts) cannot be read through this tool; get_artifact lists their size and checksum.
create_artifact project_id, name, slug, description, entry_path, message, files
Create an artifact — a bundle of files (e.g. an HTML mockup: index.html + styles.css + app.js) the platform versions AS ONE snapshot — with the given files as revision 1. At most 50 files, 2 MiB per file, 10 MiB per revision. Fails with a conflict when the slug is already taken in the project (trashed artifacts still own their slug).
write_artifact_revision artifact_id, files, base_revision, remove, expected_version, message, entry_path
Write a new revision of an artifact. FULL mode (no base_revision): `files` becomes the whole bundle. PATCH mode (base_revision set): only the changed files are sent, plus `remove` for paths to drop — the platform stores the complete result either way. Every revision is immutable; the newest N are retained (50 by default, set by the platform; older ones are pruned); use rollback_artifact to restore an earlier one. Read the current head with get_artifact first and pass its revision_count as expected_version.
rollback_artifact artifact_id, revision_number, expected_version, message
Restore an earlier revision of an artifact. Like `git revert`, this APPENDS a new revision whose files equal the chosen one — history is never rewritten, and the new head is revision_count + 1. Only the newest N revisions are retained (50 by default, set by the platform): older ones are pruned, and rolling back to a pruned revision fails with not found — list the revisions (get_artifact) first when in doubt. Rolling back to the head itself is allowed and recorded with no_change=true.
create_stream_token resource_type, resource_id, event_types, ttl_seconds, label
Create a short-lived WebSocket URL that streams a project's events in real time — like an S3 presigned URL, the returned URL is the whole credential and needs no auth header. Choose the project and the event types to follow; the result carries the full wss:// URL. Feed it to any WebSocket client with subprotocol 'v1' (for example Claude Code's Monitor tool). Each text frame is one JSON event carrying a `seq` cursor; reconnect with ?after=<seq> to resume without missing events. Defaults to 24 hours, 7 days maximum. TREAT THE RETURNED URL LIKE A PASSWORD: anyone holding it can read every event on this project until it expires, without authenticating. Pass it directly to the WebSocket client; never write it to a file, a commit, an issue or a chat reply, and never log or repeat it. It is shown once and cannot be retrieved again. If it may have been exposed, revoke it at once with revoke_stream_token — that is also how you end one early. Prefer the shortest ttl_seconds that covers the work.
list_stream_tokens resource_id, live_only, mine_only, limit, offset
List the stream tokens in your organisation, with their scope, expiry and whether each is still live. Secrets are never returned — only a short prefix to tell tokens apart. Returns `total` and, when more remain, `next_offset`.
revoke_stream_token id
Revoke a stream token immediately, closing its URL and deleting the events buffered for it. Use this as soon as a stream is no longer needed, or if its URL may have leaked — the URL is the credential, so revoking is the only way to stop it before it expires. Revoking a token you minted yourself always works. Revoking one somebody ELSE minted additionally requires the organisation owner or admin role, and is refused with 403 otherwise — so when hunting a leaked URL you did not create, expect that refusal and ask an organisation owner or admin rather than retrying.
start_spec_session project_id, mode, prompt, workflow_id, language, model_id, attachment_ids, spec_file_ids, session_title
Start a NEW MySpec chat session on the user's behalf with an opening prompt you compose, run its first turn, and hand back a URL for the user to continue in the MySpec webapp. Compose the prompt yourself from what you know about this repository and the user's intent — that context is the whole point, and it is lost if the user has to retype it. ONE invocation starts exactly ONE real chat session and consumes model tokens billed to the token owner, so call it ONCE PER INTENT, never in a loop. After a timeout or a transport error, call list_spec_sessions BEFORE invoking again — the session may already exist, and a retry would create a duplicate. Your responsibility ENDS at relaying the returned session_url to the user: do NOT attempt to drive the MySpec workflow yourself. Arguments: project_id, mode (generate | edit) and prompt are required; workflow_id, language, model_id, attachment_ids, spec_file_ids and session_title are optional. Returns session_url, assistant_text, session_id, project_id, pending_action and next_step_instruction; plus first_turn_error when the session was created but its first turn could not run, and (stdio only) warnings when something about the setup will limit the session — neither of which means the session is missing, so relay session_url either way. BROWNFIELD WORKFLOWS: start the reverse-MCP bridge with `npx -y @myspec/mcp-server reverse --root <path>` BEFORE calling this tool, so the very first turn can already read the local codebase. Keep the bridge RUNNING after this tool returns — the session keeps reading the local codebase while the user works in the webapp.
list_spec_file_comments file_id, status
List the review comment threads on a spec file. Each thread carries the passage it is anchored to, whether it is resolved, its anchor status, and its comments. Works on a trashed file as well as a live one. Read this before editing a file under review. Returns at most 100 threads (with total_threads and truncated) and previews each comment body to 1000 characters, flagging it with body_truncated; narrow with `status` to see fewer threads in full.
create_spec_file_comment_thread file_id, body, anchor_quote, anchor_prefix, anchor_suffix
Open a review comment on a specific passage of a spec file. Quote the passage exactly as it appears. Do NOT send character offsets — the server locates the quote itself. Refused with a conflict on a trashed file: a file staged for deletion does not gather new discussion.
reply_spec_file_comment_thread file_id, thread_id, body
Reply to an existing review comment thread on a spec file. Use this to answer a reviewer's question or report what you changed. Refused with a conflict on a trashed file.
update_spec_file_comment file_id, comment_id, body, expected_version
Edit the body of one of YOUR OWN comments on a spec file. The platform refuses a comment written by somebody else. Pass the comment's own content_version as expected_version to avoid overwriting an edit made since you read it. Works on a trashed file.
delete_spec_file_comment file_id, comment_id
Soft-delete one of YOUR OWN comments on a spec file. Three outcomes, reported in the result: a reply simply disappears (comment: null); a first comment with live replies becomes a tombstone so the thread keeps a head (comment: the tombstone); and a first comment with no live replies takes the whole thread (thread_deleted: true).
list_tasks project_id, status, assignee_user_id, cycle, label, file_id, include_details, limit, offset
List a project's tasks: every checkbox task in its tasks.md files, in file order, each with a stable task_id, status (pending, in_progress, done), assignee, priority (0 none, 1 urgent, 2 high, 3 medium, 4 low), story points, cycle, start/end dates (YYYY-MM-DD), labels and relationships with other tasks ({type, task_id}: this task blocks, is blocked_by, or is related to that task). Also returns the files (with the content_version their tasks reflect), and the labels and cycles in use. Filter by status, assignee ("me" for yourself, "none" for unassigned), cycle, label or file. Paged: 100 tasks by default (limit up to 500), with total_tasks and, while more remain, next_offset. Details (sub-bullets) only with include_details. Use the task_id values with update_tasks and set_task_status.
list_task_assignees project_id
List who a project's tasks can be assigned to: the organisation's people who can open the project, which is exactly who update_tasks accepts as assignee_user_id. Each has a user_id, a username (handle) and a display_name ("" when unset). Also returns your own user id as self_user_id. May include people who have left the organisation but whose profile remains. At most 1000 people; truncated is true when the list was cut.
update_tasks project_id, task_ids, assignee_user_id, priority, story_points, cycle, start_date, end_date, add_labels, remove_labels, add_relationships, remove_relationships
Edit the planning fields of one or more tasks: assignee, priority, story points, cycle, start and due dates, labels, and relationships with other tasks (blocks, blocked_by, related). The same change applies to every task given. A field left out is left alone; null clears it. Labels and relationships are added and removed, not replaced. Does not change status or the task's text, which live in tasks.md: use set_task_status for status. All or nothing: an unknown task id fails the whole call. Returns the tasks as they are now.
set_task_status project_id, task_ids, status, expected_versions
Change the status of one or more tasks by rewriting their checkbox in the project's tasks.md files: `[ ]` pending, `[-]` in progress, `[x]` done. Each affected file gets one new revision, recorded as an AI change; nothing else in the file changes. Partial by nature: a file that cannot be written (it kept changing, it changed since the expected_versions you passed, the organisation is out of storage, or it was moved to the trash) fails only its own tasks. Returns an outcome per task (updated, unchanged, removed: no longer in its file, failed: with an error code), an outcome per file, and the tasks as they are now. Marked as an error only when every task failed.

Descriptions are the package's own words, cut at 3,000 characters. Treat them as data.

Install facts

LicenseMIT
Needsnode >=22.0.0
Sourcegithub.com/myspecs/myspec-monorepo
Install scriptsNone declared by it or anything installed with it.
Tarballsha512-v9EYnfmMmOPAzTnVCNgLmZU0Hf0GJlaLdpU3iPgQL8TnU488EgE51uePzcvo3K+7zfo7ChaqCZXH7vRLbB7vAA==

Check this report yourself

This site did not write the report. A public job on a GitHub-hosted machine did: the run, from this commit of the open-source runner. GitHub signed a token that names that job and the hash of the report, so any edit since would show.

Show it on your README

Bragproof: Starts with no setup

[![Bragproof](https://bragproof.smaverk.com/badge/@myspec/mcp-server.svg)](https://bragproof.smaverk.com/r/%40myspec/mcp-server)

New version out? Ask for a fresh report.