mcptools
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@mcptools search the web for latest AI news"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
web-tools-mcp-server (mcptools)
An all-in-one MCP server for local LLMs (Gemma, gpt-oss, Qwen, etc. via LM Studio, Ollama, or any MCP client). It gives a model the capabilities small local models lack: live web access, precise math, the current date, shell + code execution, filesystem and SQLite access, a persistent memory, and semantic search (RAG) over your own files using LM Studio embeddings.
Transport: stdio (local).
Config:
~/.config/mcptools/config.json(auto-created on first run).No cloud keys required — search is Startpage (+ Google headless fallback), weather is open-meteo, embeddings are your local LM Studio.
39 tools in 10 config-gated groups.
Table of contents
Related MCP server: MCP Server
Tools (39, in 10 groups)
Group | Tools | Notes |
web |
| Mojeek; HTML→text/markdown |
compute |
| mathjs; IANA timezones |
system |
| unrestricted, timeout-guarded; python/node/bash |
files |
| grep/glob/exact + line-based edits without shell quoting; PDF→JSON; binary-safe downloads |
data |
| any local |
knowledge |
| keyless |
feeds |
| keyless live sources |
memory |
| persistent SQLite at |
tasks |
| persistent checklist at |
rag |
| embeddings via LM Studio; vectors in |
⚠️ Security:
run_command,execute_code,write_file,edit_file,download_fileand unrestricted filesystem access run with your user's full privileges (only timeouts, no sandbox — as configured). Run this server only with clients/models you trust, or disable those groups in the config (see Configuration).
Tool reference
Every tool returns human-readable text plus structuredContent for programmatic
clients. Responses are capped to limits.characterLimit and truncated with a note.
web
Tool | Key arguments | Returns |
|
| Ranked results: title, url, snippet |
|
| Cleaned page content (scripts/nav stripped) |
|
| Absolute links with anchor text |
|
| title, description, canonical, OpenGraph |
|
| status, headers, body (any REST API) |
compute
Tool | Key arguments | Returns |
|
| mathjs result (functions, constants, units) |
|
| ISO + human time + unix seconds |
system
Tool | Key arguments | Returns |
|
| File contents, pageable with line numbers (PDFs auto-detected, text extracted per page) |
|
| Bytes written (creates parent dirs) |
|
| Entries with type + size |
|
| exit code, stdout, stderr (system shell) |
|
| exit code, stdout, stderr |
|
| Desktop notification via |
| — | mcptools version, enabled/disabled groups, config + data paths |
files
Tool | Key arguments | Returns |
|
| Matches as |
|
| Matching paths with sizes |
|
| Exact in-place replacement (errors if not found / ambiguous); returns edited region |
|
| Line-based edits (use |
|
| PDF as JSON: metadata + per-page text (write to file for big PDFs, then |
|
| Binary-safe download to disk (pairs with |
data
Tool | Key arguments | Returns |
|
| Rows (SELECT) or change count |
|
| Matching slice as JSON |
knowledge
Tool | Key arguments | Returns |
|
| Article title, summary, url |
|
| Current conditions + daily forecast |
feeds
Tool | Key arguments | Returns |
|
| Feed items: title, link, date, summary (RSS 2.0 + Atom) |
|
| Stories with points, comments, discussion link (Algolia, keyless) |
|
| Papers with abstract + |
|
| Full caption transcript as text |
memory
Tool | Key arguments | Returns |
|
| Confirmation (upsert) |
|
| Matching notes |
| — | All keys with tags + timestamps |
|
| Confirmation |
tasks
Tool | Key arguments | Returns |
|
| New task id |
|
| Checklist with ids |
|
| Marks done (or deletes with |
rag
Tool | Key arguments | Returns |
|
| Files + chunks indexed (skips on backend error) |
|
| Ranked passages with source + score |
| (none) | Indexed sources with chunk counts + store totals |
|
| Chunks deleted + updated store totals |
Install & build
cd ~/Code/mcptools
npm install
npm run build # compiles to dist/Requires Node ≥ 22 (uses the built-in node:sqlite). python3 is needed
for execute_code with language: "python".
Run
node dist/index.js # speaks MCP over stdioThe first run writes a default ~/.config/mcptools/config.json.
Register with a client
Install script (recommended)
npm run install:claude # Claude Code (~/.claude.json, user scope)
npm run install:opencode # OpenCode (~/.config/opencode/opencode.json)
npm run install:all # bothMerges the mcptools entry into the client's user-level config (everything
else in the file is preserved; re-running updates the entry in place). Build
first — the script points the client at dist/index.js.
Claude Code (manual)
claude mcp add mcptools -- node ./dist/index.jsOpenCode
Add to ~/.config/opencode/config.json (global) or a project opencode.json,
under the mcp key:
{
"mcp": {
"mcptools": {
"type": "local",
"command": ["node", "./dist/index.js"],
"enabled": true
}
}
}Restart OpenCode (or start a new session) to spawn the server. Verify with
/mcp inside OpenCode, or list tools from the TUI.
Generic MCP client (mcpServers JSON)
{
"mcpServers": {
"mcptools": {
"command": "node",
"args": ["./dist/index.js"]
}
}
}💡 Small local models choose tools more reliably with fewer of them. If 39 tools is too many for your model, disable groups in the config rather than the client — see below.
Configuration (~/.config/mcptools/config.json)
{
"groups": { // turn whole tool groups on/off
"web": true, "compute": true, "system": true, "data": true,
"knowledge": true, "memory": true, "rag": true
},
"tools": { // per-tool overrides win over group, e.g.:
// "run_command": false,
// "write_file": false
},
"limits": {
"timeoutMs": 30000, // network/exec timeout
"characterLimit": 25000, // max chars per tool response
"maxOutputBytes": 1000000 // max bytes captured from a process/file/fetch
},
"lmstudio": {
"baseUrl": "http://localhost:1234/v1",
"embeddingModel": "text-embedding-nomic-embed-text-v1.5"
},
"paths": { "dataDir": "~/.config/mcptools" }
}Per-tool overrides (
tools) take precedence over the group switch.Hot env overrides:
MCPTOOLS_LMSTUDIO_URL,MCPTOOLS_EMBED_MODEL.For a safe, read-only-ish profile set
"system": false(and optionally"data": false) ingroups.
A copy of the defaults lives in config.example.json.
RAG / embedding server
rag_index and rag_search call any OpenAI-compatible
/v1/embeddings endpoint. At startup, the server auto-detects which
embedding backend is running and uses it:
The URL in
lmstudio.baseUrl(orMCPTOOLS_LMSTUDIO_URLenv var) — primary.http://localhost:1234/v1— LM Studio.http://localhost:8089/v1— highllama / llama.cpp.
For each, it hits /v1/models with a 2 s timeout, picks the most promising
embedding model (configured id, then anything with embed/bge/minilm/e5
in its id), and verifies with a real probe /v1/embeddings call that the
model actually returns a numeric vector. This rejects llama.cpp servers that
are currently serving a chat model (the common case — llama.cpp loads one
model at a time, so if it's busy with gemma-4-26B for chat, it can't also
serve embeddings). For single-model servers with a name that doesn't match
the patterns, the same probe is used as a fallback so custom-named embedding
models are still discovered. The chosen endpoint is logged to stderr on
startup, e.g.:
[mcptools/rag] using LM Studio (1234) with model "text-embedding-nomic-embed-text-v1.5"If nothing is reachable, the configured values are used and the embedding call will fail with a clear error.
⚠️ Embedding model crashing on GPU — load it on CPU
Observed with text-embedding-nomic-embed-text-v1.5: the model crashes
mid-request ("The model has crashed... Exit code: null") and then fails to reload
("unable to allocate CUDA0 buffer"). Root cause is GPU/CUDA, not the input —
the same chunks embed fine on CPU. This tiny (~84 MB) model runs fast on CPU.
Load it CPU-only (the lms CLI is the only place that sets the GPU layer-offload
ratio — the REST API does not):
~/.lmstudio/bin/lms unload --all
~/.lmstudio/bin/lms load text-embedding-nomic-embed-text-v1.5 --gpu off -yLoads in ~0.5 s and embeds reliably. An explicitly-loaded model has no TTL and stays resident until LM Studio restarts, so JIT won't reload it on the GPU.
Persisting it (headless / no GUI). The GUI "per-model default" is the correct knob and applies to JIT loading, but LM Studio only writes it from the GUI (into an undocumented internal file — don't hand-edit it). Headless options:
Run the
lms load ... --gpu offcommand at login (e.g. asystemd --useroneshot unit, or append it to your existing~/.lmstudio/load-optimized.sh).Or just run it once per LM Studio session before using RAG.
If you have the GUI on another machine: My Models → ⚙️ → set GPU offload = 0.
Defensive hardening (any backend)
Inputs are embedded one at a time (batched array input crashes some backends).
Each embed call retries with backoff.
rag_indexskips any chunk that still fails and reports the count, instead of aborting the whole index.
Development
npm run dev # tsx watch (no build step)
npm run build # type-check + compile
npm run clean # remove dist/Project layout:
src/
index.ts entry: load config, register enabled tools, stdio
config.ts config load/merge + defaults (~/.config/mcptools)
registry.ts group/tool enable gating
types.ts shared types
services/ search (mojeek), lmstudio (embeddings), vectorstore
tools/ web, compute, system, data, knowledge, memory, rag
util/ http, html, exec, sqlite, format, errorsData files created at runtime under paths.dataDir (default
~/.config/mcptools): config.json, memory.db, rag.db.
Available Tools
39 toolsarxiv_searcharXiv SearchARead-onlyIdempotent
Search arXiv for scientific papers (no API key). To read a paper: take pdf_url from a result, download_file it, then read_file the PDF.
Args:
query (string): Search terms, e.g. "mixture of experts quantization".
sort ('relevance'|'latest'): Ranking (default 'relevance').
limit (number): Max papers (default 5).
Returns papers with title, authors, date, abstract and pdf_url.
Example: { "query": "speculative decoding", "sort": "latest" }
Inputs: query (required), sort, limit.
calculatorCalculatorARead-onlyIdempotent
Evaluate a mathematical expression precisely. Use this for ANY arithmetic instead of computing in your head.
Supports +, -, *, /, ^, parentheses, functions (sqrt, sin, cos, log, etc.), constants (pi, e), and units (e.g. "5 km in miles", "2 GB in bytes").
Args:
expression (string): The expression to evaluate.
Returns the result as a string.
Examples:
{ "expression": "(1234 * 5678) / 90" }
{ "expression": "sqrt(2) * 100" }
{ "expression": "120 km/h in m/s" }
Inputs: expression (required).
current_datetimeCurrent Date and TimeARead-only
Return the current date and time. Use this whenever you need to know "now", compute ages/durations, or timestamp something — never guess the date.
Args:
timezone (string): Optional IANA timezone, e.g. "Europe/Madrid", "America/New_York", "UTC". Defaults to the system timezone.
Returns the ISO timestamp, a human-readable string, and component parts.
Example: { "timezone": "Asia/Tokyo" }
Inputs: timezone.
download_fileDownload FileADestructiveIdempotent
Download a URL to a local file, binary-safe (PDFs, images, archives...). Use fetch_url for reading web pages as text; use this when you need the actual file on disk — e.g. download a PDF, then read_file it to extract the text.
Args:
url (string): The file URL (http/https).
path (string): Destination file path. Parent directories are created.
Returns the saved path, size and content type.
Example: { "url": "https://arxiv.org/pdf/1706.03762", "path": "~/papers/attention.pdf" }
Inputs: url (required), path (required).
edit_fileEdit FileADestructive
Replace an exact string in a text file. Much safer than rewriting the whole file with write_file — the rest of the file is untouched. The old_string must match exactly (including whitespace) and, unless replace_all is true, must appear exactly once; include surrounding lines to make it unique. If you cannot reproduce the text exactly, use edit_lines instead.
Args:
path (string): File to edit.
old_string (string): Exact text to find.
new_string (string): Replacement text.
replace_all (boolean): Replace every occurrence (default false).
dry_run (boolean): Preview the change without writing (default false).
Returns a confirmation plus the edited region with line numbers, so you can verify without re-reading the file.
Example: { "path": "~/app/config.py", "old_string": "DEBUG = True", "new_string": "DEBUG = False" }
Inputs: path (required), old_string (required), new_string (required), replace_all, +1 more.
edit_linesEdit LinesADestructive
Edit a file by line numbers: replace a line range, insert text, or delete lines. Use read_file with line_numbers=true first to get the exact numbers. Prefer edit_file when you can copy the exact text; use this when you know the location but not the exact characters. IMPORTANT: line numbers shift after every edit — re-read the file before making another line-based edit.
Args:
path (string): File to edit.
mode ('replace'|'insert_before'|'insert_after'|'delete'): What to do (default 'replace').
start_line (number): First line affected (1-based, inclusive).
end_line (number): Last line affected (default start_line). Ignored for inserts.
new_text (string): The new lines (required except for delete).
dry_run (boolean): Preview the change without writing (default false).
Returns a confirmation plus the edited region with line numbers.
Example: { "path": "~/app.py", "mode": "replace", "start_line": 12, "end_line": 14, "new_text": "def main():\n run()" }
Inputs: path (required), start_line (required), mode, end_line, +2 more.
execute_codeExecute CodeADestructive
Run a snippet of code and return its output. Use this to compute, transform data, or test logic instead of doing it in your head.
Args:
language ('python'|'node'|'bash'): Interpreter to use.
code (string): The source code. Print results to stdout.
stdin (string): Optional text piped to the program's stdin.
Returns exit code plus captured stdout/stderr. Killed after the configured timeout.
Example: { "language": "python", "code": "print(sum(range(100)))" }
Inputs: language (required), code (required), stdin.
extract_linksExtract Links from URLARead-onlyIdempotent
Fetch a page and list all outbound hyperlinks (absolute URLs with their anchor text).
Useful for crawling, finding a specific subpage, or mapping a site section.
Args:
url (string): The page to scan.
limit (number): Max links to return, 1-200 (default 50).
Returns a list of { text, href }.
Example: { "url": "https://news.ycombinator.com", "limit": 30 }
Inputs: url (required), limit.
fetch_urlFetch URL as TextARead-onlyIdempotent
Download a web page and return its readable content as clean text or Markdown (scripts, styles, nav and ads stripped).
Use after web_search to actually read a page, or whenever you have a URL whose content you need.
Args:
url (string): The http(s) URL to fetch.
format ('text' | 'markdown'): 'text' (default) for plain readable text, 'markdown' to preserve headings/links/lists.
Returns the page content. Long pages are truncated; fetch a more specific URL if needed.
Example: { "url": "https://example.com/article", "format": "markdown" }
Inputs: url (required), format.
find_filesFind FilesARead-onlyIdempotent
Find files by name across a directory tree using a glob pattern. Use this instead of walking directories one list_directory call at a time. Skips .git and node_modules.
Args:
path (string): Root directory to search (default ".").
pattern (string): Glob, e.g. ".pdf", "**/.test.ts", "config". Matched against the path relative to root.
max_results (number): Max files returned (default 100).
Returns matching paths with sizes.
Example: { "path": "~/Documents", "pattern": "*.pdf" }
Inputs: pattern (required), path, max_results.
get_page_metadataGet Page MetadataARead-onlyIdempotent
Fetch a page and return its metadata: title, description, canonical URL and OpenGraph tags. Cheaper than fetching full content when you only need to identify a page.
Args:
url (string): The page URL.
Returns { title, description, canonical, og }.
Example: { "url": "https://example.com" }
Inputs: url (required).
hackernewsHacker News SearchARead-onlyIdempotent
Search Hacker News stories (Algolia API, no key). Great for tech news, library/tool opinions and discussions.
Args:
query (string): Search terms. Empty string returns the current front page.
sort ('relevance'|'date'): Ranking (default 'relevance').
limit (number): Max stories (default 10).
Returns stories with points, comment count and discussion link.
Example: { "query": "llama.cpp", "sort": "date" }
Inputs: query, sort, limit.
http_requestHTTP RequestA
Make a raw HTTP request to any API and return status, headers and body. Use for REST/JSON APIs that need GET/POST/PUT/DELETE with custom headers or a body.
Args:
url (string): Target URL.
method ('GET'|'POST'|'PUT'|'PATCH'|'DELETE'): default 'GET'.
headers (object): Optional header map, e.g. { "Authorization": "Bearer ..." }.
body (string): Optional request body (send JSON as a stringified object).
Returns { status, statusText, headers, body }.
Example: { "url": "https://api.github.com/repos/nodejs/node", "headers": { "Accept": "application/vnd.github+json" } }
Inputs: url (required), method, headers, body.
json_queryQuery JSON/CSV FileARead-onlyIdempotent
Extract a slice of a local JSON or CSV file with a path query, instead of reading the whole file into context. CSV files are parsed using the first row as headers (every row becomes an object).
Path syntax: "." is the whole document, ".key.subkey" descends into objects, "[0]" indexes an array, "[]" maps over every element. Examples: ".users[0].name", ".items[].price", ".results[]" .
Args:
path (string): The .json or .csv file.
query (string): Path query (default "." — but prefer something narrower).
limit (number): If the result is an array, max elements returned (default 100).
Returns the matching slice as JSON.
Example: { "path": "~/data/users.json", "query": ".users[].email" }
Inputs: path (required), query, limit.
list_directoryList DirectoryARead-onlyIdempotent
List entries in a directory with type (file/dir) and size.
Args:
path (string): Directory path (default current working directory).
Returns a list of { name, type, size_bytes }.
Example: { "path": "~/project/src" }
Inputs: path.
memory_deleteDelete MemoryADestructiveIdempotent
Delete a saved memory by key.
Args:
key (string): The key to delete.
Example: { "key": "user_name" }
Inputs: key (required).
memory_listList Memory KeysARead-onlyIdempotent
List all saved memory keys (with tags and timestamps) without their full values. Use to see what you have stored.
Args: none.
Returns a list of { key, tags, updated_at }.
Inputs: no named parameters.
memory_recallRecall MemoryARead-onlyIdempotent
Retrieve saved notes. Give an exact key for a single note, or a search term to match across keys/values/tags. Call this when you might already know something the user told you earlier.
Args:
key (string): Optional exact key to fetch.
search (string): Optional text to match (substring, case-insensitive).
limit (number): Max results for a search (default 10).
Returns matching { key, value, tags, updated_at } records.
Example: { "search": "timezone" }
Inputs: key, search, limit.
memory_saveSave MemoryAIdempotent
Persist a fact/note under a key so you can recall it in future turns or sessions. Overwrites an existing key.
Args:
key (string): Short unique identifier, e.g. "user_timezone".
value (string): The information to remember.
tags (string): Optional comma-separated tags for later filtering.
Example: { "key": "user_name", "value": "Sergio", "tags": "profile" }
Inputs: key (required), value (required), tags.
notifyDesktop NotificationA
Show a desktop notification to the user (via notify-send). Use to announce that a long task finished or that you need attention.
Args:
title (string): Notification title.
message (string): Notification body.
urgency ('low'|'normal'|'critical'): Urgency level (default 'normal').
Example: { "title": "Done", "message": "Finished indexing 240 files." }
Inputs: title (required), message (required), urgency.
pdf_to_jsonConvert PDF to JSONARead-onlyIdempotent
Convert a local PDF into structured JSON: document metadata (title, author...), page count, and the text of each page. Use read_file for a quick plain-text read; use this when you need page structure, metadata, or a JSON file other tools can consume (e.g. json_query).
Args:
path (string): The PDF file.
output_path (string): Optional .json destination. If set, the full JSON is written there and only a summary is returned — best for large PDFs.
Returns { source, total_pages, metadata, pages: [{ page, text }] }.
Example: { "path": "/papers/attention.pdf", "output_path": "/papers/attention.json" }
Inputs: path (required), output_path.
rag_deleteDelete a RAG SourceADestructiveIdempotent
Remove all indexed chunks for a source from the RAG store. The source must match the path stored at index time (an absolute path) — use rag_list to find exact paths. The original file on disk is not touched.
Args:
source (string): Source path to delete, exactly as shown by rag_list.
Returns the number of chunks deleted and the updated store totals.
Example: { "source": "/tmp/ragcorpus/cooking.md" }
Inputs: source (required).
rag_indexIndex Documents for RAGAIdempotent
Embed and index a local file or directory into the semantic search store so you can retrieve relevant passages later with rag_search. Re-indexing the same source replaces its old chunks.
Requires an embedding server (LM Studio or llama.cpp) running with an embedding model loaded.
Args:
path (string): File or directory to index. Directories are walked for text/code files.
max_files (number): Cap on files when indexing a directory, 1-2000 (default 200).
Returns the number of files and chunks indexed.
Example: { "path": "~/project/docs" }
Inputs: path (required), max_files.
rag_listList Indexed RAG SourcesARead-onlyIdempotent
List the source files currently indexed in the RAG store, with the chunk count for each. Use this to see what is indexed and to get the exact source paths needed by rag_delete.
Returns a list of { source, chunks } and overall store totals.
Example: {}
Inputs: no named parameters.
rag_searchSemantic Search (RAG)ARead-onlyIdempotent
Search previously indexed documents by meaning and return the most relevant passages with their source files. Use this to ground answers in the user's own files/docs that you indexed with rag_index.
Requires an embedding server (LM Studio or llama.cpp) running with an embedding model loaded.
Args:
query (string): What to look for (a question or topic).
top_k (number): Number of passages to return, 1-20 (default 5).
Returns ranked { source, score, text } passages.
Example: { "query": "how is authentication handled", "top_k": 5 }
Inputs: query (required), top_k.
read_fileRead FileARead-onlyIdempotent
Read a file from the local filesystem. Text files are returned as-is (UTF-8); PDF files are detected automatically and their text is extracted page by page.
Args:
path (string): Absolute or ~-relative file path.
offset (number): Start reading at this 1-based line (default 1). Use with max_lines to page through big files.
max_lines (number): Optional cap on lines returned.
line_numbers (boolean): Prefix each line with "N: " (default false). Use before edit_lines, so you know the exact line numbers.
Returns the file contents. Use list_directory first if unsure of the path.
Example: { "path": "/project/README.md" } or { "path": "/app.py", "offset": 100, "max_lines": 50, "line_numbers": true }
Inputs: path (required), offset, max_lines, line_numbers.
read_rssRead RSS/Atom FeedARead-onlyIdempotent
Fetch and parse an RSS or Atom feed, returning the latest items. Use for news sites, blogs, podcasts and release feeds — cheaper and more current than web_search when you know the feed.
Args:
url (string): The feed URL (e.g. "https://news.ycombinator.com/rss").
limit (number): Max items returned (default 10).
Returns items as { title, link, date, summary }.
Example: { "url": "https://lwn.net/headlines/rss", "limit": 5 }
Inputs: url (required), limit.
run_commandRun Shell CommandADestructive
Run a shell command and return its stdout, stderr and exit code. Runs through the system shell, so pipes, redirects and globs work. Killed after the configured timeout.
WARNING: This executes arbitrary commands on the local machine. Only use commands the user would approve.
Args:
command (string): The shell command line to run.
cwd (string): Optional working directory.
Returns exit code plus captured stdout/stderr.
Example: { "command": "ls -la | head", "cwd": "~/project" }
Inputs: command (required), cwd.
search_filesSearch FilesARead-onlyIdempotent
Search file contents under a directory with a regular expression (like grep -rn, but no shell quoting to get wrong). Skips binary files, .git and node_modules. Use this to find where something is defined or mentioned.
Args:
path (string): Directory (or single file) to search.
pattern (string): Regular expression, e.g. "function \w+Tools". Plain text works too.
glob (string): Optional filename filter, e.g. ".ts" or "src/**/.js".
context_lines (number): Lines of context around each match (default 0).
max_results (number): Max matching lines returned (default 50).
Returns matches as file:line: text, grouped by file.
Example: { "path": "~/project", "pattern": "registerTool", "glob": "*.ts" }
Inputs: path (required), pattern (required), glob, context_lines, +1 more.
server_infoServer InfoARead-onlyIdempotent
Report the mcptools server version and configuration. Call this when asked what version of mcptools is running, or which tool groups are enabled.
Args: none.
Returns { name, version, node, enabled_groups, disabled_groups, tool_overrides, config_path, data_dir }.
Inputs: no named parameters.
sqlite_querySQLite QueryADestructive
Run a SQL statement against a local SQLite database file and return the rows or change count. Use this to query or modify structured data instead of inventing values.
To explore an unfamiliar database, first run: SELECT name FROM sqlite_master WHERE type='table'.
Args:
db_path (string): Path to the .sqlite/.db file (created if missing unless read_only).
sql (string): A single SQL statement. SELECT/PRAGMA return rows; INSERT/UPDATE/DELETE return a change count.
params (array): Optional positional parameters for ? placeholders.
read_only (boolean): Open read-only (default false).
Returns rows (for SELECT) or { changes, lastInsertRowid }.
Example: { "db_path": "~/data/app.db", "sql": "SELECT * FROM users WHERE age > ?", "params": [18] }
Inputs: db_path (required), sql (required), params, read_only.
todo_addAdd TodoA
Add a task to the persistent todo list. When given a multi-step job, add each step first, then work through them with todo_done — this keeps you from forgetting steps.
Args:
text (string): The task description.
Returns the new task id.
Example: { "text": "Write tests for the parser" }
Inputs: text (required).
todo_doneComplete TodoADestructiveIdempotent
Mark a todo as done by id (from todo_list). Set remove=true to delete it entirely instead.
Args:
id (number): The task id.
remove (boolean): Delete the task instead of marking done (default false).
Example: { "id": 3 }
Inputs: id (required), remove.
todo_listList TodosARead-onlyIdempotent
List tasks on the todo list (pending by default). Check this at the start of a work session to see what is still open.
Args:
include_done (boolean): Also show completed tasks (default false).
Returns tasks as a checklist with ids.
Inputs: include_done.
weatherWeatherARead-onlyIdempotent
Get the current weather and a short forecast for a place (via open-meteo, no API key). Use this for any "what's the weather" question — never guess.
Args:
location (string): City or place name, e.g. "Madrid", "Tokyo, Japan".
days (number): Forecast days, 1-7 (default 3).
Returns current conditions plus a daily min/max forecast.
Example: { "location": "Barcelona", "days": 3 }
Inputs: location (required), days.
web_searchWeb SearchARead-onlyIdempotent
Search the web and return ranked results (title, URL, snippet).
Use this whenever you need current information, facts you are unsure about, or to find pages to read with fetch_url. No API key required.
Args:
query (string): What to search for. Plain keywords work best.
limit (number): Max results, 1-20 (default 5).
response_format ('markdown' | 'json'): Output style (default 'markdown').
Returns a list of { title, url, snippet }. Follow up with fetch_url on a result's url to read it.
Example: { "query": "rust async runtime comparison", "limit": 5 }
Inputs: query (required), limit, response_format.
wikipedia_lookupWikipedia LookupARead-onlyIdempotent
Look up a topic on Wikipedia and return a concise summary (with the page URL). Prefer this over web_search for encyclopedic facts about people, places, concepts and events.
Args:
query (string): The topic or article title to look up.
lang (string): Wikipedia language code (default 'en').
Returns { title, summary, url }.
Example: { "query": "Alan Turing" }
Inputs: query (required), lang.
write_fileWrite FileADestructive
Write text to a file, creating parent directories as needed. Overwrites by default; set append=true to append.
Args:
path (string): Destination file path.
content (string): Text to write.
append (boolean): Append instead of overwrite (default false).
Returns a confirmation with bytes written.
Example: { "path": "~/notes.txt", "content": "hello", "append": true }
Inputs: path (required), content (required), append.
youtube_transcriptYouTube TranscriptARead-onlyIdempotent
Fetch the transcript (captions) of a YouTube video as plain text. Use to summarize or answer questions about a video without watching it. Fails cleanly if the video has no captions.
Args:
video (string): Video URL or 11-character video ID.
lang (string): Preferred caption language code (default 'en'; falls back to whatever exists).
Returns the transcript text.
Example: { "video": "https://www.youtube.com/watch?v=dQw4w9WgXcQ" }
Inputs: video (required), lang.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 39 tools
Each tool has a clearly distinct purpose with detailed descriptions. Even within groups like memory (save/recall/list/delete) or RAG (index/search/list/delete), the boundaries are clear. There is minimal overlap; for example, fetch_url and http_request are differentiated by use case (readable content vs raw API).
The majority of tools follow a consistent verb_noun pattern (e.g., download_file, fetch_url, list_directory). However, a few tools deviate: arxiv_search (noun_verb), hackernews (single noun), weather, calculator, json_query, pdf_to_json. These are still descriptive but break the predominant pattern.
With 39 tools, the server is on the high end of the typical range. While each tool has a place, the overall count feels heavy for a single server. However, it covers many domains (file, web, memory, RAG, todo, utilities) and is not excessive enough to be confusing.
The tool surface is comprehensive for a general-purpose assistant, covering file operations, web interactions, memory, RAG, todo lists, code execution, shell commands, SQLite, and more. Minor gaps exist (e.g., no image processing or email), but the core workflows are well-supported.
Maintenance
Related MCP Connectors
A comprehensive Model Context Protocol (MCP) server that enables AI assistants to interact with yo…
MCP server that lets your AI assistant build and deploy apps, hosting each one at a live URL.
An MCP server that gives your AI access to the source code and docs of all public github repos
Related MCP Servers
- AlicenseCqualityDmaintenanceA multi-tool MCP server that enhances local LLMs with web search, document reading, scholarly research, Wikipedia access, and calculator functions. Provides comprehensive tools for information retrieval and computation without requiring API keys by default.221MIT
- AlicenseNot gradedqualityDmaintenanceA modular MCP server providing file operations, web search, URL scraping, and sandboxed command execution for LLM interactions.1MIT
- AlicenseAqualityBmaintenanceA universal MCP server that enables any LLM or AI agent to access expert skills from your local filesystem.3214 npm42MIT
- AlicenseNot gradedqualityDmaintenanceA complete MCP server that gives LM Studio's Qwen3 (or any local LLM) full coding agent capabilities including 80+ tools for file operations, command execution, git, web search, memory, planning, and a full skills system.14MIT