Get your first cited answer

① Connect

1. Open claude.ai/customize/connectors.
2. Click Add custom connector.
3. Name it Pipeworx, paste the first URL below, and save. Leave the OAuth client ID/secret blank — the connector registers itself.
4. Click Connect on the connector and sign in with GitHub. This is the step that matters: without it you are capped at 50 calls a day, with it you get 200 — free, same connection.
5. In a new chat, open the tools menu and enable Pipeworx (you should see ~38 tools).

https://gateway.pipeworx.io/oauth/mcp

https://gateway.pipeworx.io/pipeworx-catalog/mcp

Not signing in still works — you get 50 calls a day anonymously, no account needed. When you run out, Pipeworx asks your client to sign you in rather than just failing, so you can do it at that point instead.

Free Claude accounts can’t add custom connectors — that’s an Anthropic plan gate, not a Pipeworx one.

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into Settings → Profile → personal preferences (or a Project’s custom instructions) so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

1. Custom connectors are added on the web and sync to mobile. Add Pipeworx once at claude.ai/customize/connectors (see the Web tab).
2. Open the Claude iOS app → it appears under your connectors automatically.
3. In a chat, enable Pipeworx from the tools menu.

https://gateway.pipeworx.io/oauth/mcp

https://gateway.pipeworx.io/pipeworx-catalog/mcp

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into claude.ai → Settings → Profile preferences (syncs to mobile) so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

1. Add Pipeworx once on the web at claude.ai/customize/connectors (see the Web tab) — it syncs to the Android app.
2. Open the Claude Android app → enable Pipeworx in a chat’s tools menu.

https://gateway.pipeworx.io/oauth/mcp

https://gateway.pipeworx.io/pipeworx-catalog/mcp

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into claude.ai → Settings → Profile preferences (syncs to mobile) so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

Option A — UI: Settings → Connectors → Add custom connector → paste the URL.

https://gateway.pipeworx.io/oauth/mcp

https://gateway.pipeworx.io/pipeworx-catalog/mcp

Option B — config file (claude_desktop_config.json):

{
  "mcpServers": {
    "pipeworx": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://gateway.pipeworx.io/oauth/mcp"
      ]
    }
  }
}

No account? Swap in https://gateway.pipeworx.io/pipeworx-catalog/mcp — same connection, 50 calls a day instead of 200.

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into Settings → Profile preferences (same account as web) so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

Easiest — one line, then restart. (Or just tell Claude Code: “add the Pipeworx MCP server at https://gateway.pipeworx.io/oauth/mcp” and it runs this for you.) Then run /mcp and authenticate: free, and it takes you from 50 calls a day to 200.

claude mcp add --transport http pipeworx https://gateway.pipeworx.io/oauth/mcp

Or install the plugin — same tools, plus an auto-routing skill so Claude reaches for ask_pipeworx without being told:

/plugin marketplace add pipeworx-io/claude-code-plugin
/plugin install pipeworx

② So it reaches for live data on its own

✓ Automatic — this install ships a routing rule, so the model prefers Pipeworx with no extra step.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

No Developer mode needed — build a Custom GPT with a Pipeworx Action:

1. ChatGPT → Explore GPTs → + Create → Configure. Name it Pipeworx.
2. Actions → Create new action → Authentication: None.
3. Schema → Import from URL → paste the OpenAPI URL below.
4. Set Privacy policy to https://pipeworx.io/privacy, then Create.

https://gateway.pipeworx.io/openapi.json

Want the full MCP connector (all tools, in the main chat)? Settings → Connectors → Developer mode → add an MCP server with https://gateway.pipeworx.io/pipeworx-catalog/mcp.

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into the Custom GPT’s Instructions field so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

1. Create the Pipeworx Custom GPT once on the web (see the Web tab).
2. In the ChatGPT app, open it from your GPTs list and chat — the Action works on mobile too.

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into the Custom GPT’s Instructions field (set on web) so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

OpenAI’s visual workflow builder (platform.openai.com/agent-builder) connects to Pipeworx as an MCP tool — no code, no auth:

1. Open Agent Builder and create (or open) a workflow.
2. Click your Agent node → Tools → + → MCP Server → + Server.
3. Paste the URL below. Label it pipeworx. Authentication: None.
4. Pick the tools to expose (or leave all selected) and Add.

https://gateway.pipeworx.io/oauth/mcp

https://gateway.pipeworx.io/pipeworx-catalog/mcp

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into the Agent node’s Instructions field so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

Add to ~/.codex/config.toml:

[mcp_servers.pipeworx]
url = "https://gateway.pipeworx.io/oauth/mcp"

No account? Swap in https://gateway.pipeworx.io/pipeworx-catalog/mcp — same connection, 50 calls a day instead of 200.

Note: run codex login first — Codex won’t load any MCP server until you’re authenticated.

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into a repo <code>AGENTS.md</code> file so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

Settings → Connectors → Add connector → Custom MCP server. Expand Advanced to reach the auth setting.

Prefer no account, or the sign-in step doesn't appear? Use https://gateway.pipeworx.io/pipeworx-catalog/mcp with Auth: None — 50 calls a day per IP, no signup.

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into a Perplexity Space → custom instructions so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

1. Add the connector once on the web (see the Web tab) — it syncs to mobile.
2. In the app, use it inside a Space that has the routing instructions.

https://gateway.pipeworx.io/oauth/mcp

https://gateway.pipeworx.io/pipeworx-catalog/mcp

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into a Perplexity Space → custom instructions so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

Mistral Le Chat → Settings → Connectors → add a custom MCP server.

https://gateway.pipeworx.io/oauth/mcp

https://gateway.pipeworx.io/pipeworx-catalog/mcp

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into an Agent / custom instructions so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

Add the connector on the web (see Web tab); it syncs to the Le Chat mobile app.

https://gateway.pipeworx.io/oauth/mcp

https://gateway.pipeworx.io/pipeworx-catalog/mcp

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into an Agent / custom instructions so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

Official extension — one command (or just ask Gemini CLI to run it):

gemini extensions install https://github.com/pipeworx-io/gemini-cli-extension

Note: Google is folding Gemini CLI into Antigravity CLI (mid-2026); the same extension/MCP mechanism carries over. The consumer Gemini web app does not support custom connectors — use the CLI or Gemini Enterprise.

② So it reaches for live data on its own

✓ Automatic — this install ships a routing rule, so the model prefers Pipeworx with no extra step.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

For orgs on Gemini Enterprise, an admin wires Pipeworx as a custom MCP server in the Google Cloud console:

1. Google Cloud console → Gemini Enterprise → Data stores → Create → Custom MCP server.
2. MCP server URL: paste the URL below. Create the data store.
3. Open it → enable the tools/actions (none are on by default).

https://gateway.pipeworx.io/pipeworx-catalog/mcp

The consumer Gemini web app (gemini.google.com) can’t add custom connectors — this is the only Gemini web/app path, and it’s admin-gated.

② So it reaches for live data on its own

✓ Automatic — this install ships a routing rule, so the model prefers Pipeworx with no extra step.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

Grok supports custom MCP connectors ("Bring Your Own MCP") across web and mobile:

1. Open grok.com/connectors → New Connector → Custom.
2. Paste the first URL below and sign in with GitHub when prompted — free, and it takes you from 50 calls a day to 200. The second URL needs no account and works at 50.
3. Save — Grok discovers the ~38 tools. It syncs to the iOS/Android apps under the same account.

https://gateway.pipeworx.io/oauth/mcp

https://gateway.pipeworx.io/pipeworx-catalog/mcp

Custom connectors may require a paid Grok plan (SuperGrok/Premium) — that’s an xAI gate, not a Pipeworx one.

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into Grok’s custom instructions / personalization settings so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

Grok Build CLI supports bring-your-own remote MCP. Add Pipeworx as a remote MCP server in its connector config:

https://gateway.pipeworx.io/oauth/mcp

https://gateway.pipeworx.io/pipeworx-catalog/mcp

See xAI’s remote-MCP docs for the exact CLI config.

② So it reaches for live data on its own

✓ Automatic — this install ships a routing rule, so the model prefers Pipeworx with no extra step.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

DeepSeek is a model, not an MCP host — its app has no custom-connector slot. To use DeepSeek with Pipeworx, select a DeepSeek model inside an MCP-capable client and point that client at Pipeworx:

1. Pick a client that lets you choose your model and add MCP servers — e.g. Cline, Continue.dev, or Cursor (each has its own quickstart here).
2. Set its model/provider to DeepSeek (DeepSeek’s API is OpenAI/Anthropic-compatible).
3. Add Pipeworx via that client using the URL below.

https://gateway.pipeworx.io/oauth/mcp

https://gateway.pipeworx.io/pipeworx-catalog/mcp

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into whichever client you run DeepSeek in (see its quickstart) so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

One-click deeplink — signs you in, 200 calls/day:

cursor://anysphere.cursor-deeplink/mcp/install?name=pipeworx&config=eyJ1cmwiOiJodHRwczovL2dhdGV3YXkucGlwZXdvcnguaW8vb2F1dGgvbWNwIn0=

Or the no-account deeplink — 50 calls/day:

cursor://anysphere.cursor-deeplink/mcp/install?name=pipeworx&config=eyJ1cmwiOiJodHRwczovL2dhdGV3YXkucGlwZXdvcnguaW8vcGlwZXdvcngtY2F0YWxvZy9tY3AifQ==

Or add to .cursor/mcp.json:

{
  "mcpServers": {
    "pipeworx": {
      "url": "https://gateway.pipeworx.io/oauth/mcp"
    }
  }
}

No account? Swap in https://gateway.pipeworx.io/pipeworx-catalog/mcp — same connection, 50 calls a day instead of 200.

② So it reaches for live data on its own

✓ Automatic — this install ships a routing rule, so the model prefers Pipeworx with no extra step.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

Windsurf is now Devin Desktop (devin.ai/download) — same Cascade panel, same config path. Edit ~/.codeium/windsurf/mcp_config.json, save, then refresh Cascade’s MCP list:

{
  "mcpServers": {
    "pipeworx": {
      "serverUrl": "https://gateway.pipeworx.io/oauth/mcp"
    }
  }
}

No account? Swap in https://gateway.pipeworx.io/pipeworx-catalog/mcp — same connection, 50 calls a day instead of 200.

Via the UI: Settings → Cascade → Configuration → MCP servers → switch to the Cascade marketplace (not Devin Local) → Add custom MCP → paste the above → save.

② So it reaches for live data on its own

✓ Automatic — this install ships a routing rule, so the model prefers Pipeworx with no extra step.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

Easiest — one line (adds to your user profile):

code --add-mcp '{"name":"pipeworx","type":"http","url":"https://gateway.pipeworx.io/oauth/mcp"}'

No account? Swap in https://gateway.pipeworx.io/pipeworx-catalog/mcp — same connection, 50 calls a day instead of 200.

Or per-project — add .vscode/mcp.json:

{
  "servers": {
    "pipeworx": {
      "type": "http",
      "url": "https://gateway.pipeworx.io/oauth/mcp"
    }
  }
}

No account? Swap in https://gateway.pipeworx.io/pipeworx-catalog/mcp — same connection, 50 calls a day instead of 200.

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into a <code>.github/copilot-instructions.md</code> file so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

Add to settings.json under context_servers:

{
  "context_servers": {
    "pipeworx": {
      "source": "custom",
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://gateway.pipeworx.io/oauth/mcp"
      ]
    }
  }
}

No account? Swap in https://gateway.pipeworx.io/pipeworx-catalog/mcp — same connection, 50 calls a day instead of 200.

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into a project <code>.rules</code> file so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

1. Settings → Tools → AI Assistant → Model Context Protocol (MCP) → Add.
2. Choose As command and enter:

npx -y mcp-remote https://gateway.pipeworx.io/oauth/mcp

No account? Swap in https://gateway.pipeworx.io/pipeworx-catalog/mcp — same connection, 50 calls a day instead of 200.

Builds with an SSE/HTTP server option can take the URL directly instead.

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into a project guidelines file so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

Add to ~/.continue/config.yaml:

mcpServers:
  - name: pipeworx
    type: streamable-http
    url: https://gateway.pipeworx.io/oauth/mcp

No account? Swap in https://gateway.pipeworx.io/pipeworx-catalog/mcp — same connection, 50 calls a day instead of 200.

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into a Continue rules block so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

MCP Servers → Remote Servers → paste the URL (Cline can’t self-add — use the panel):

https://gateway.pipeworx.io/oauth/mcp

https://gateway.pipeworx.io/pipeworx-catalog/mcp

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into a <code>.clinerules</code> file so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

Moonshot’s Kimi Code speaks native HTTP MCP. Edit ~/.kimi-code/mcp.json (or run /mcp-config in the TUI) — a url entry is an HTTP server:

{
  "mcpServers": {
    "pipeworx": {
      "url": "https://gateway.pipeworx.io/oauth/mcp"
    }
  }
}

Verify with /mcp in the session. Kimi Code is BYO-model — run kimi login or kimi provider first. For 200/day add a headers object with X-API-Key.

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into KIMI.md / the agent’s system prompt so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

OpenClaw speaks native HTTP MCP. Add Pipeworx in one command:

openclaw mcp add pipeworx --url https://gateway.pipeworx.io/oauth/mcp --transport streamable-http

Or edit openclaw.json under mcp.servers. Verify with openclaw mcp doctor --probe. For 200/day, add --header X-API-Key "YOUR_KEY".

{
  "mcp": {
    "servers": {
      "pipeworx": {
        "url": "https://gateway.pipeworx.io/oauth/mcp",
        "transport": "streamable-http"
      }
    }
  }
}

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into your OpenClaw agent system prompt (openclaw.json) so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

Hermes Agent (Nous Research) speaks native HTTP MCP. Add Pipeworx in one command:

hermes mcp add pipeworx --url https://gateway.pipeworx.io/oauth/mcp

Verify with hermes mcp test pipeworx. Or edit ~/.hermes/config.yaml under mcp_servers. For 200/day, add a headers entry with your key:

mcp_servers:
  pipeworx:
    url: "https://gateway.pipeworx.io/oauth/mcp"
    headers:
      X-API-Key: "YOUR_KEY"

② So it reaches for live data on its own

Models answer “known” facts from memory unless told otherwise. Paste this into a HERMES.md file / the agent’s system prompt so it reaches for live data on its own:

When I ask for current or factual data — economic stats, SEC/EDGAR filings, stock or crypto prices, weather, FDA drug info, patents, clinical trials, or government records — use the Pipeworx connector (ask_pipeworx) instead of answering from memory or web search. Prefer structured Pipeworx data with citations.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

One command writes the MCP config into your local AI tools (Claude Code, Codex, Cursor):

npx pipeworx use ask-pipeworx

Scope to a single source, or target a specific client:

npx pipeworx use edgar
npx pipeworx use ask-pipeworx --for=cursor

Browse and test packs from the terminal:

npx pipeworx list
npx pipeworx search weather
npx pipeworx test edgar

② So it reaches for live data on its own

✓ Automatic — this install ships a routing rule, so the model prefers Pipeworx with no extra step.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.

① Connect

Calling from your own code — one import, ask anything or call a specific tool:

npm install @pipeworx/sdk

import { Pipeworx } from '@pipeworx/sdk';

const px = new Pipeworx();
const r = await px.ask('What is the US trade deficit with China?');
console.log(r.data, r.toolSelected);

② So it reaches for live data on its own

✓ Automatic — this install ships a routing rule, so the model prefers Pipeworx with no extra step.

③ Ask your first question

This is the part that matters. Paste it in:

Using Pipeworx, what was the US unemployment rate last month?

You should get last month’s actual rate as a number, with the name of the series it came from and the period it covers — enough to go and check it yourself. That’s a cited answer. Ask it anything else you’d normally have to look up, or start with a recipe.