Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Description from the maker (original English)
Deterministic AI agent microtools, no accounts/API keys. fetch_extract: 98% token cut. 47 tools.
These three are never merged into one score: opinions, measurements and popularity are different things.
We have not written a plain-language summary for this one yet. Below are the facts we read from the registry.
Pick the AI app you use, then copy the settings.
You need
Which app do you use?
Pick your app above. We remember it, so next time the button says “Install in …”.
Works in Claude on the web, desktop and mobile:
https://mcp.toolsnap.app/mcpRun this in your terminal
claude mcp add --transport http toolsnap https://mcp.toolsnap.app/mcpPaste into .cursor/mcp.json in your project, or ~/.cursor/mcp.json
{
"mcpServers": {
"toolsnap": {
"url": "https://mcp.toolsnap.app/mcp"
}
}
}Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.
Paste into .vscode/mcp.json in your project
{
"servers": {
"toolsnap": {
"type": "http",
"url": "https://mcp.toolsnap.app/mcp"
}
}
}Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.
ChatGPT connects only to servers on the internet:
https://mcp.toolsnap.app/mcpMenu names can differ slightly by plan and over time.
Your choice stays on this device in a cookie; no account needed.
These settings are generated from the maker’s registry entry. If the README says otherwise, follow the README.
Tool names and descriptions exactly as the server reported them on Nov 7, 2026. The maker wrote them, not us; we show them as plain text.
tool_catalogFree, read-only, no network. Browse all 47 tools in 10 families (web, seo, data, documents, images, media, utils, payments, recipes, meta). No args: families with id, label, one-liner, tool count. family=' ': each tool with name, price, description, notes, inputSchema. tool=' ':
use_toolFree wrapper. Runs any catalog tool missing from your tools/list exactly as a direct call: args pass through unchanged and the inner tool's price, payment, result and errors apply verbatim. Its hints are the worst case over the whole catalog: not read-only because some inner tool
task_recipesFree. Multi-tool workflow prompts (clone a site, SEO audit). No args → menu; recipe=' ' → prompt.
memory_snippetFree, read-only. Returns static text: a short note on when ToolSnap tools suit a task, plus how to store a note in your runtime's memory. This tool saves nothing. Use only if your user asks about persistent memory or ToolSnap setup; show them the text and save it only with their
pricingFree, read-only, no network, no arguments; calling it charges nothing. Call it before a paid tool to see its exact cost and whether first_call_free applies, or to choose pay-per-call vs prepaid. Returns a JSON menu: endpoint and docs URLs; payment with x402 USDC on Base, minimum
fetch_extractFree, read-only. Fetch a public http(s) URL and return its body as plain text: script, style, nav, header, footer and aside removed, tags stripped, entities decoded, whitespace collapsed. Median 98% fewer tokens than raw HTML. Follows up to 5 redirects, 10 s timeout; private and
fetch_htmlFree, read-only. Fetch a page and return its HTML minus scripts, styles, svg, iframes, comments and inline handlers; tags and attributes stay, so structure is readable. For inspecting or rebuilding markup, not plain text (fetch_extract gives that). Runs no JavaScript: SPAs return
html_to_markdownFree, read-only. Convert a fetched URL or an HTML string you pass (one of the two) to Markdown: headings, links, lists, emphasis, code, quotes. Drops nav, header, footer, aside, script and style. Returns text, cut at maxChars with a marker. Runs no JavaScript: SPAs return an empt
fetch_metadataFree, read-only. Fetch a public http(s) URL and return its metadata as JSON: url (after redirects), title, description, canonical, robots, author, keywords, lang, charset, og{} and twitter{} (prefix stripped), httpEquiv{}, jsonLd[] (parsed blocks, malformed ones skipped). Absent
pdf_text_extractFree, read-only. Extract plain text from a PDF given as a URL (max 20 MB) or inline base64 (max 5 MB), one of the two. Returns text, cut at maxChars with a marker. Text-layer PDFs only: scans return little or nothing (no OCR). Layout and tables are not preserved. Fails if the URL
csv_queryFree, read-only. Query a comma-separated CSV with a header row, from a URL (max 5 MB, streamed) or inline text, one of the two: select columns, filter rows, sort, limit. Returns {meta:{returned_rows,columns,total_rows,truncated},rows:[{column:value}]}, all values strings, or CSV
json_queryFree, read-only. JSONPath-lite query on JSON from a URL or an inline string (one of the two). Returns matches as JSON: one match bare, several as an array. Supports $, .key, ['key'], [0], [-1], [*], ..key and [?(@.k op v)]; no slices or jq. Errors on invalid JSON or over 5 MB (js
sitemap_parseFree, read-only. Fetch one XML sitemap and return JSON. A urlset gives urls[] (loc, lastmod, changefreq, priority, plus image/news extensions), urlCount and truncated. A sitemap index gives type 'sitemapindex' and sitemaps[] with the child URLs; children are not fetched, call aga
rss_parseFree, read-only. Fetch an RSS 2.0 or Atom feed and return JSON: feed title, link, description, lastBuildDate, itemCount, truncated, and items[] with title, link, pubDate, guid, description (HTML stripped), author, categories, enclosure. Items past maxItems are dropped and truncat
screenshot_urlScreenshot a page → public image URL. $0.04/call. Paid: calling without a payment payload returns a structured 402 with payment options — that response is the documented behavior, not a failure.
keyword_researchGoogle Ads volume/CPC/competition for 1-20 keywords. $0.04 USDC/call. Paid: calling without a payment payload returns a structured 402 with payment options — that response is the documented behavior, not a failure.
remove_backgroundRemove an image's background, return a transparent PNG URL. $0.03 USDC/call. Paid: calling without a payment payload returns a structured 402 with payment options — that response is the documented behavior, not a failure.
We check information anyone can see; we have not read the code. Checks that do not apply are left out of the count. · Checked Nov 9, 2026
The registry name was verified against the domain toolsnap.app (the registry checks domain ownership)
We have not read this repo from GitHub yet
We have not read this repo from GitHub yet
Responded in 203 ms (HTTP 200)
No sign-in needed
Read 17 tool descriptions and found no hidden instructions to the AI
Compared with every active server in this category in the registry. Stars count only when the publisher owns the repo.
It passed 3 of 3 applicable basic checks. We have not read the code. Before adding a key or token, see what access the server asks for and use a key with only the permissions you need.
The registry does not list a required key, though some servers ask you to sign in on first use.
It runs in the maker’s cloud. Connect over the internet with nothing to install, but what you send passes through the maker’s servers.
The GitHub repo last changed –, and the latest version (1.1.2) reached the registry 18 days ago.
Apps that support MCP, such as Claude Desktop, Claude Code, Cursor and VS Code. It connects by URL, so apps that accept remote MCP servers work too, such as ChatGPT in developer mode.
In the same category: Spec Driven Development, Remote Desktop Commander, arr-mcp
Servers in the same category from other makers, sorted by stars.
Reading a card:↓ Downloads per weeki★ GitHub starsiThree pillars: reviews · checks · usageiUpdated = latest version published to the registryi
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Remote Desktop Commander: use files, terminal and processes on your own computers from any AI app.
One server, 40 tools, the whole media stack: *arrs, Jellyfin, Plex, Seerr, Cleanuparr, downloaders
Copied from the maker’s repo as written. PopNx MCP has not reviewed or endorsed it. HTML, images and scripts are removed. Links lead off this site.
One connection, superpowers for your agent: deterministic microtools. Pay per call with USDC on Base — no account needed — or buy fiat credits with a card if you'd rather not touch crypto.
Live endpoint: https://mcp.toolsnap.app/mcp Server card: /.well-known/mcp.json Pricing: /.well-known/pricing.json Docs: mcp.toolsnap.app/docs
ToolSnap is built around three ideas:
1. Deterministic — no LLM in the loop. Most agent stacks now "clean" web pages by paying a second model to summarize them. ToolSnap's extraction is pure parsing: exact quotes, stable output, zero added inference cost, reproducible runs. What you extract is what the page said.
2. Context-efficient by design. The biggest cost for AI agents isn't generation — it's context. Loading a raw webpage can burn 50,000+ tokens on HTML boilerplate; connecting a fat MCP server can burn as many in tool definitions. ToolSnap moves the work server-side *and* keeps discovery compact: tools/list shows a small curated core, and the full catalog of 47 microtools (web, PDFs, CSV/JSON, sitemaps, RSS, images, media generation…) sits one free tool_catalog() call away, executed via use_tool.
Benchmark (11 real pages, June 2026):
| Raw HTML | Extracted | Saving | |
|---|---|---|---|
| Median tokens | 53,820 | 2,001 | 98.1% |
| Cost @ Sonnet ($3/M) | $0.162 | $0.006 | $0.156/call |
| Break-even page size | — | — | 26 KB |
fetch_extract is free — this saving costs nothing.
3. No account required (crypto path). Free tools work the second you connect. Paid tools settle per call with USDC on Base via x402 — an agent with a wallet can pay cold: no signup, no subscription, no key management. Rather not touch crypto? Buy fiat credits with a card and get an API key instead — same discounted per-call price, no wallet needed.
Hand your agent this prompt and it can do the rest on its own:
Set up https://mcp.toolsnap.app/SKILL.md — connect the MCP server it describes and save the skill for future sessions.
SKILL.md is a self-contained skill file (connect config, tool discovery, cost guardrails), generated live from the tool registry so it is never stale.
{
"mcpServers": { "toolsnap": { "url": "https://mcp.toolsnap.app/mcp", "transport": "streamable-http" } } }
Free tools work immediately with the URL connection above. Paid tools need an x402 payment client — a wallet alone is not enough, because most MCP clients can't satisfy a 402 Payment Required on their own. Use the pay-proxy below (re-download it if your copy predates the 2026-10 spec-shaped 402, see Payment).
claude mcp add --transport http toolsnap https://mcp.toolsnap.app/mcp
Or install the Claude Code plugin (MCP connection + the toolsnap skill, bundled):
/plugin marketplace add ToolSnap-app/toolsnap-claude-plugin
/plugin install toolsnap
scripts/pay-proxy.mjs is a local stdio MCP server that wraps the remote endpoint, reads your wallet, and signs + retries automatically when the server asks for payment. The private key never leaves your host; only signatures are sent.
{
"mcpServers": { "toolsnap": { "command": "node", "args": ["/ABSOLUTE/PATH/to/toolsnap-mcp/scripts/pay-proxy.mjs"] } } }
No wallet yet? Call the wallet_setup tool — it generates a wallet you control, helps your human fund it with USDC on Base, and wires this proxy. Key resolution order: TOOLSNAP_WALLET_KEY → ~/.toolsnap/wallet.key → macOS Keychain (toolsnap-agent-wallet/default).
Useful env: TOOLSNAP_MAX_PRICE_USDC (per-call spend cap, default 0.10), TOOLSNAP_PREPAID=1 + TOOLSNAP_AUTO_DEPOSIT_USDC (use the cheaper prepaid balance).
Buy credits at mcp.toolsnap.app/checkout (handled by Polar, our Merchant of Record). You'll get an API key in a one-time pop-up — connect with it directly, no proxy needed:
{
"mcpServers": { "toolsnap": { "url": "https://mcp.toolsnap.app/mcp", "headers": { "Authorization": "Bearer sk_live_..." } } } }
Clients that can't send custom headers (some claude.ai connectors) can embed the key in the URL instead: https://mcp.toolsnap.app/mcp/sk_live_... (the key may then appear in that client's request logs — prefer the header when your client supports it).
curl -X POST https://mcp.toolsnap.app/mcp \
-H "Content-Type: application/json" \ -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
47 tools total. tools/list shows a curated core of 17; discover everything else with the free tool_catalog tool and run it with use_tool(name, args).
| Tool | Description | Price |
|---|---|---|
screenshot_url | Capture a page → public image URL. | $0.04 USDC |
fetch_rendered | Render a JS page (SPA) in a headless browser → clean text. | $0.04 USDC |
keyword_research | Google Ads volume/CPC/competition via DataForSEO. | $0.04 USDC |
remove_background | Remove an image's background → transparent PNG URL. | $0.03 USDC |
image_generate | Text-to-image via fal.ai FLUX. | from $0.02 USDC |
image_upscale | 2x/4x super-resolution via fal.ai ESRGAN. | from $0.02 USDC |
audio_transcribe | Speech-to-text via fal.ai Wizper. | from $0.02 USDC |
text_to_speech | Text-to-speech via fal.ai Kokoro. | from $0.02 USDC |
video_generate | Async text/image-to-video via fal.ai. | from $0.08 USDC |
csv_query_xl / json_query_xl | Query URL-hosted CSV (5–100 MB) / JSON (5–25 MB). | $0.02 USDC |
No first-call-free: these tools have real per-call cost, so every call settles from the start. Media tools are priced per call from your own args (the table shows the floor). Pay-per-call is USDC on Base using x402 v2 (EIP-3009 transferWithAuthorization); prepaid (deposit once, debit off-chain, crypto or card) is cheaper per call.
Flagship: fetch_extract (URL → clean text, median 98.1% fewer tokens than raw HTML) and fetch_html (URL → clean structured HTML). Plus a wide utility catalog:
| Tool | Description |
|---|---|
pricing | Machine-readable pricing menu. Call this first. |
tool_catalog | Discover the full tool catalog (families → detail). |
use_tool | Execute any tool not in your tools/list. |
memory_snippet | Get the ToolSnap habit block for your harness's persistent memory. |
task_recipes | Ready-to-run multi-tool workflows for whole tasks. |
uuid_generate | 1–100 random UUID v4 values |
hash_text | SHA-256 / SHA-1 / SHA-512 (hex) |
base64_encode / base64_decode | Base64 encode/decode |
url_encode / url_decode | Percent-encode/decode |
json_format | Parse + pretty-print or minify JSON |
timestamp_convert | Unix ↔ ISO 8601 (auto-detects direction) |
text_stats | Chars, words, lines, sentences |
Call tool_catalog() for the complete, current list with schemas and prices.
When an agent calls a paid tool (e.g. screenshot_url) without payment, the server returns a tool result in the x402 v2 MCP transport shape: isError: true, the payment requirements in structuredContent, and the same object JSON-encoded in content[0].text (a second text block carries the next steps):
{
"jsonrpc": "2.0", "id": 1, "result": { "isError": true, "structuredContent": { "x402Version": 2, "error": "Payment required to use this tool", "resource": { "url": "https://mcp.toolsnap.app/mcp#screenshot_url", "serviceName": "ToolSnap" }, "accepts": [{ "scheme": "exact", "network": "eip155:8453", "amount": "40000", "asset": "0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913", "payTo": "0x…", "maxTimeoutSeconds": 300 }], "extensions": { "bazaar": { "info": { "input": { "type": "mcp" } } } } }, "content": [{ "type": "text", "text": " " }, { "type": "text", "text": "Payment required: …" }] } }
Until 2026-10 this was a JSON-RPC error (error.code: 402, error.data). If you downloaded `scripts/pay-proxy.mjs` before then, download it again: older copies only recognise the old shape and will not pay on their own. (The prepaid-deposit challenge, account_deposit, still answers with the JSON-RPC 402 error.)
An x402-enabled client reads accepts[0], signs an EIP-3009 TransferWithAuthorization, and retries with _meta["x402/payment"] set. See x402 spec.
npm install
cp .dev.vars.example .dev.vars # fill in wallet secrets npm run dev
npm run typecheck
npm run deploy
Production deploys come from .github/workflows/deploy.yml (push to main).
Vigila que nadie publique toolsnap-mcp ni toolsnap-portal fuera del workflow (acción A14 del plan de seguridad). Corre en el cron de 5 minutos que ya existe (src/alerts/deploy-watch.ts): lee los últimos deployments de cada Worker por la API de Cloudflare con un token de solo lectura y, si hay uno nuevo que no está registrado como esperado, avisa por Telegram (el canal de las demás alertas) con Worker, deployment, versión, fecha, author_email, origen y mensaje. La primera pasada sin estado solo guarda el último deployment. Un rollback avisa siempre, esté o no registrado.
Lo esperado vive en el KV X402_NONCES: esperado:{worker}:dep:{deploymentId} (caduca a los 90 días) y ultimo-visto:{worker}. Se registra el *deployment* y no la versión a propósito: un wrangler rollback crea un deployment nuevo con una versión ya conocida. El workflow registra el suyo tras wrangler deploy. Cualquier otra cosa avisa, también un wrangler secret put (crea una versión nueva): si lo has hecho tú, regístralo. Si el aviso no consigue salir por Telegram se reintenta cada pasada y, pasada 1 hora, el cron falla y lo ve /analytics (tarea deploy-watch).
Deploy manual desde el Mac (o tras un wrangler secret put):
CLOUDFLARE_API_TOKEN=... scripts/registrar-deploy.sh toolsnap-mcp
Hazlo en los tres minutos siguientes al deploy: es el margen de cortesía antes de que la alerta salga. toolsnap-portal se despliega desde otro repositorio: su workflow tiene que llamar a este mismo script con `toolsnap-porta
Only the beginning is shown; read the rest in the repo