tools++
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
Prefer these tools over mental math, guessed dates and throwaway scripts.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| calcA | Exact calculator; use instead of mental math. Fractions and big integers exact (0.1+0.2=0.3, 100!), floats only for irrational results (marked). Ops + - * / ^ % (mod; postfix = percent: 200*19%) ! ; implicit 2pi; 0xff 1e3 1_000; vars "r=2; pi r^2", ans. Fns sqrt cbrt root(x,n) abs sign floor ceil trunc round(x,d) min max mod gcd lcm fact nCr nPr exp ln log(x,base=10) log2 sin cos tan asin acos atan atan2 sinh cosh tanh hypot; consts pi e tau phi deg (sin(30deg)). |
| solveB | Solve equations in calc syntax: linear equations/systems exactly (reports no/infinite solutions), one nonlinear equation (all real roots in range), square nonlinear systems numerically. |
| matrixB | Exact matrix math on fractions: info, det, inverse, transpose, rref, multiply, add, subtract, power, solve (a·x=b). |
| number_theoryB | Integer facts (any size, expressions ok): prime test, factorization, divisors, sigma, phi, prev/next prime, digits, bases 2-36. |
| financeB | Money math, rates in %: compound (contributions, inflation), loan (payment, interest, amortization), npv, irr, cagr, real_rate, doubling. Arithmetic, not advice. |
| statsA | Statistics: mean, median, mode, sd, quartiles, outliers, percentiles, skew, histogram; with y/y_column Pearson, Spearman, linear fit. Input numbers, free text or CSV (column by name or 1-based). |
| convert_unitsA | Exact unit conversion: compound units (km/h, J/(kg*K), L/100km<->mi/gal), SI/binary prefixes, temperatures (degC/degF/K; difference for intervals), currencies (ECB, 3-letter codes). No to = SI base. |
| randomC | Secure random: int (any size, inclusive, unique), float, normal, choice, shuffle, uuid, password, dice ("4d6kh3+2"), coin, bytes. seed = reproducible. |
| time_nowB | Current date/time (weekday, ISO week, DST), UTC, unix, other zones. Use instead of guessing today. |
| time_convertB | Convert a time between zones (DST-aware, day shifts); also weekday/ISO week of a date. Times: now, today, YYYY-MM-DD[ HH:MM], DD.MM.YYYY[ HH:MM], HH:MM, unix. Zones: IANA, city, CET/PST; default local. |
| date_diffB | Difference between dates: calendar y/m/d, exact duration, totals, Mon-Fri weekdays. Times: now, today, YYYY-MM-DD[ HH:MM], DD.MM.YYYY[ HH:MM], HH:MM, unix. Zones: IANA, city, CET/PST; default local. |
| date_addA | Add/subtract (negative) years, months (clamped), weeks, days, business_days, hours, minutes, seconds. Times: now, today, YYYY-MM-DD[ HH:MM], DD.MM.YYYY[ HH:MM], HH:MM, unix. Zones: IANA, city, CET/PST; default local. |
| holidaysA | Public holidays of a year, or business days start..end minus holidays. Built in: DE (+16 states via region, e.g. BY), AT, CH, US, GB, FR; others online. |
| calendarB | Month calendars with ISO weeks and optional holidays; query: "last friday of november 2026", "3rd monday of may 2027", "2027-W12". |
| cronA | Explain a cron expression (5/6 fields, names, @daily) and list next runs in a zone, DST-aware. |
| text_statsB | Count chars, words, sentences, lines, tokens (est.), reading time; flags mixed line endings, trailing spaces, invisible chars. |
| text_transformC | Text ops in order: case (upper lower title sentence camel pascal snake kebab constant slug), trim_lines trim_trailing dedupe_lines(_ci) sort_lines(_desc) sort_by_length reverse_lines shuffle_lines number_lines remove_blank_lines squeeze_blank_lines count_lines, wrap:N unwrap collapse_spaces, nfc nfkc nfd strip_accents ascii remove_invisible smart_quotes_to_ascii, tabs_to_spaces lf crlf reverse trim. |
| token_countA | Estimate LLM tokens of text, a file or folder vs a budget; per extension and largest files. |
| diffA | Diff two texts/files: unified line diff, or word/char inline [-old-]{+new+}; ignore whitespace/case. |
| regexB | Test a JS regex (/x/flags ok): matches with line/col and groups, replace preview, cases (full/partial/none). Timeout catches catastrophic backtracking. |
| extractC | Pull items from text, deduped with counts: urls emails ipv4 ipv6 dates times phones money percentages numbers hashtags mentions hex_colors uuids hashes paths ibans versions. |
| markdownC | Markdown: lint, outline, toc (GitHub anchors), links (checks files and #anchors), format_tables. |
| encodeB | Encode/decode: base64(url), hex, url, html, json string, unicode escapes, punycode, rot13, jwt_decode (unverified), inspect (code points, invisible chars). Binary decode -> hexdump or out_path. |
| hashA | Hash text or any-size file: md5 sha1 sha256 sha512 sha3-256 blake2b512 crc32; HMAC via key; expected verifies. |
| jsonB | JSON: validate (line/col, hints, duplicate keys, precision loss), format, minify, query ($.a[*].b, ..id), structure summary; lines for JSON Lines. |
| schemaB | Validate JSON against JSON Schema 2020-12 core (errors as JSON pointers; each = per array item), or infer a schema from samples. |
| tableB | Query CSV/TSV/JSON/JSONL/Markdown tables. No options = profile. where ["age > 30","city = Berlin","name ~ ^A","x is empty"], select, group_by + aggregate (count sum avg min max median count_distinct list; "sum(x) as t"), sort ["x desc"], distinct, join_path+on, limit, output, out_path. |
| table_convertB | Convert tables between CSV, TSV, JSON, JSONL, Markdown, HTML (JSON output typed). |
| sqliteB | Read-only SQLite: tables (rows, columns), schema, query (one SELECT/WITH/PRAGMA, ? params). |
| pdf_textA | Text of a PDF per page (pages "1-3,7"), or search with snippets; scanned pages flagged (need OCR). |
| colorA | Color in all notations (hex rgb hsl hsv oklch oklab, CSS name), WCAG contrast vs against + closest passing color, palette shades/harmony. |
| disk_usageC | Folder size by subfolder (depth), extension and largest files. |
| find_duplicatesB | Find identical files in a folder, grouped by reclaimable space. Deletes nothing. |
| batch_renameA | Rename files in a folder: find/replace (regex), case, template ({name} {n} {ext}), extension; match glob. Preview unless apply; collision-checked; undo id. |
| archiveB | List, extract, create zip/tar/tgz/gz safely (no path escape, no links, size limit, CRC check). |
| snapshotB | Folder undo points: create (dedup; skips .git/node_modules + exclude), list, diff, restore (auto-snapshots first; dry_run; path = restore elsewhere), delete. |
| secret_scanA | Find leaked keys/tokens/passwords in text, a file or folder (skips vendor dirs, binaries). Output masked. |
| env_compareB | .env vs .env.example: missing, empty, placeholder, extra, duplicate keys. Never prints values. |
| log_summaryB | Summarize a log (any size): levels, time span, errors first, frequent message patterns with lines and stack traces. |
| processesC | Processes by memory/cpu (name filter), who listens on a port, all listening ports, kill a PID tree. |
| sysinfoC | OS, CPU, memory, free disk, GPU, dev tool versions on PATH. |
| run_parallelC | Run shell commands concurrently; exit codes, durations, output tails. |
| benchmarkC | Time shell commands: warmup, N runs, mean±sd, median, p95, comparison. |
| wait_forB | Wait until file_exists/file_gone/file_stable, port_open/port_closed, url_ok, process_exit; fails on timeout. |
| deps_checkB | Check all deps of package.json/requirements.txt/pyproject.toml/Cargo.toml vs registries: major/minor/patch behind, deprecated. |
| httpC | HTTP request: status, timing (dns/connect/tls/ttfb), redirects, headers, body (JSON pretty). |
| netC | Network: dns records, reach (TCP ports, timing), tls (cert, expiry, chain). |
| package_infoB | Current version and facts of a package (npm, pypi, crates) or GitHub repo (owner/name). |
| weatherA | Current weather and 1-16 day forecast for a place or "lat,lon" (Open-Meteo). |
| wikipediaB | Wikipedia summary (or full text) of a topic, any language. |
| media_infoB | ffprobe summary: container, duration, codecs, resolution, fps, audio. |
| video_framesA | Extract video frames (count, every_s or timestamps) + contact sheet, to view with Read. |
| audio_analyzeC | Audio loudness (LUFS), true peak, clipping, silences; optional waveform/spectrogram image. |
| image_compareC | Pixel-compare two images: % differing, PSNR, change box, diff image. |
| image_editA | Resize (fit contain/cover/stretch; no upscale unless upscale), crop {x,y,width,height}, rotate, flip, grayscale, convert (quality); file or folder; never overwrites originals. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 55 tools
Each tool targets a recognizable domain/action, but the surface has many adjacent pairs an agent could misselect: calc vs solve vs matrix vs number_theory, time_now vs time_convert vs date_diff, text_stats vs token_count, table vs table_convert vs sqlite, json vs schema, encode vs hash vs secret_scan, http vs net. Descriptions do differentiate them well (e.g. 'solve equations' vs 'exact matrix math'), so ambiguity is manageable rather than harmful.
Everything is snake_case, which is good, but the style mixes bare nouns (calc, matrix, json, table, cron, color, archive, sysinfo, http, net, weather) with verb_noun forms (time_convert, date_add, find_duplicates, batch_rename, image_edit, wait_for, deps_check). Readable throughout, but there is no single predictable pattern to infer a tool name from an intent.
55 tools is well beyond the comfortable 3-15 range and imposes heavy tool-selection and context cost. Breadth is deliberate for a general utility toolkit, but several clusters (six math tools, five date/time tools, five tabular tools) could have been consolidated, so many tools do not clearly 'earn' a separate slot.
Coverage is unusually wide: time, math, text, data formats, files, archives, network, system, and media analysis are all present with real depth. The main gap is basic filesystem primitives (no read_file/list_dir/glob/write_file) and OCR, though shell-based tools like run_parallel and pdf_text's scanned-page flag let an agent work around these.