bob-research-tools

Paid web research, fact-check, page summary and page extraction tools for AI agents.

CommunityOnlineNo sign-inGlobalFreeRead-only

What it can do

  • Research.Answer: Answers one open research question from the live web in at most 80 words with up to 3 source URLs, returned as {answer, sources}, usually in 4 to 9 seconds. Use it for a synthesized a
  • Research.Factcheck: Checks one factual claim against the live web and returns {verdict, explanation, sources}, where verdict is supported, contradicted or unclear, usually in 4 to 9 seconds. Write the
  • Page.Summarize: Condenses one public web page into an 80-word summary with up to 4 key points, returned as {url, summary, key_points}, usually in 3 to 6 seconds. The page is fetched live, so the summa

What data it sees

Do you need an account

No: the server works without sign-in

Paid web research, fact-check, page summary and page extraction tools for AI agents. Pay per call in USDC with x402, Base or Solana. A failed call is not charged.

How to connect

How to connect bob-research-tools to Claude

The server address to paste into Claude settings:

https://bob-research-tools--bob-olsi.run.tools

  1. Open Claude (claude.ai or the desktop app).
  2. Go to Settings → Connectors.
  3. Click Add custom connector.
  4. Paste the server address copied below into Remote MCP server URL and click Add.
  5. In a chat, click + → Connectors and switch the new connector on.

Custom connectors are available on Free, Pro, Max, Team and Enterprise plans (Free is limited to one). On Team and Enterprise an organization Owner adds the connector first under Organization settings → Connectors.

Authorization

After you click Add, a sign-in window for the service opens. Sign in with your own account and approve access. Claude never sees your password.

How to connect bob-research-tools to ChatGPT

The server address to paste into ChatGPT settings:

https://bob-research-tools--bob-olsi.run.tools

  1. Open ChatGPT in a browser (chatgpt.com). A Plus, Pro, Business, Enterprise or Edu plan is required.
  2. Turn on developer mode once: Settings → Apps → Advanced settings → Developer mode.
  3. Open Settings → Connectors and click Create.
  4. Fill in the form: Name (anything), Description (one line about what the service does), MCP server URL (copy it below).
  5. Under Authentication choose OAuth if the service requires sign-in, otherwise None. Click Create.
  6. In a new chat open + → Apps/Connectors and enable the connector.

OpenAI has renamed this section before (Connectors → Apps/Plugins). If the label differs, search settings for "developer mode". On Business/Enterprise workspaces an admin must allow custom connectors first.

Authorization

On first use ChatGPT opens the service's sign-in window. Sign in and approve access.

How to connect bob-research-tools to Cursor

The server address to paste into Cursor settings:

https://bob-research-tools--bob-olsi.run.tools

Fastest: click Open in Cursor below and confirm the prompt.

Manually:

  1. In Cursor open Settings → Cursor Settings → MCP and click Add new global MCP server.
  2. Paste the JSON copied below into ~/.cursor/mcp.json (per project: .cursor/mcp.json in the repo root).
  3. Save the file. The server appears in the MCP list; authorize it there if asked.

Authorization

If the service needs sign-in, an authorize button appears next to the server in the MCP list.

Server tool list (8)

Raw names from tools/list. Only developers need these.

research.answerAnswers one open research question from the live web in at most 80 words with up to 3 source URLs, returned as {answer, sources}, usually in 4 to 9 seconds. Use it for a synthesized answer; use web.search for raw links or research.factcheck to verify a claim. Name the people, places and dates the question depends on, and ask one question per call: several in one call get one blended answer. Read-only. Paid per call with x402, up to 60 calls a minute per client; a call that fails, times out or hits capacity is not charged.
research.factcheckChecks one factual claim against the live web and returns {verdict, explanation, sources}, where verdict is supported, contradicted or unclear, usually in 4 to 9 seconds. Write the claim as a plain statement such as 'The Eiffel Tower is in Berlin' and split a compound claim into separate calls. Use research.answer for open questions. Read-only. Paid per call with x402; a failed call is not charged.
page.summarizeCondenses one public web page into an 80-word summary with up to 4 key points, returned as {url, summary, key_points}, usually in 3 to 6 seconds. The page is fetched live, so the summary reflects it now; pages behind a login or paywall may return little. Pass the address of the specific page, not a site's home page. Use page.extract for the page's own text. Read-only. Paid per call with x402; if the page cannot be read in time the call fails and is not charged.
web.searchRuns a web search and returns ranked results as {query, results: [{title, url, snippet}]}. Use it for raw links to read yourself; use research.answer for a written answer. Plain phrases work as well as keywords; since and until (YYYY-MM-DD) limit results to a publish-date range, which suits recent news. Needs an x402 payment per call; if the search fails or finds nothing, you are not charged.
page.extractReturns the main text of one public web page as clean markdown, unchanged and not summarized, up to 8,000 characters, as {url, markdown, truncated}, usually in 2 to 5 seconds. Pass the address of the specific page, not a site's home page; pages that need a login cannot be read, and longer pages are cut and flagged truncated. Use page.summarize for a short version. Read-only. Paid per call with x402, up to 60 calls a minute per client; a failed call is not charged.
web.newsReturns recent news on one topic, person, company or event as {query, window_days, summary, developments: [{headline, date, source}]}: a summary of about 50 words and up to 4 dated developments, newest first, usually in 3 to 7 seconds. A short topic or name works better than a full question; days sets the look-back from today (1 to 30, default 7), so a quiet topic needs a larger value. Use web.search for raw links. Read-only; only what the dated results support is returned. Paid per call with x402; if nothing recent is found the call fails and is not charged.
research.marketBuilds an evidence brief for a prediction-market question (Kalshi, Polymarket or similar) from the live web: {question, as_of, summary, evidence_for, evidence_against, key_dates, resolution_notes, sources}, usually in 5 to 8 seconds. Pass the question as the market words it, including its date or threshold (for example 'by Dec 31, 2026' or 'above 4.5%'). It reports evidence and what decides resolution; it gives no probability, forecast or trading advice. Use research.answer for general questions. Read-only. Paid per call with x402; if it cannot answer in time the call fails and is not charged.
research.companyReturns a short sourced brief on one company as {company, summary, what_it_does, facts, recent, sources}, usually in 5 to 8 seconds. Pass a website address such as stripe.com to avoid name clashes; a bare name resolves to the best-known company with that name and the summary says when the name is ambiguous. Facts such as founding year, location, size, funding or leadership appear only when the sources state them, and conflicting figures are reported as the most recent one with its date. Read-only. Paid per call with x402; if the company cannot be described in time the call fails and is not charged.