The Model Context Protocol is an open standard that lets an AI assistant call a data source directly instead of guessing from what it read during training. You add one URL to your assistant and it can then query our database while it answers you.

Which assistants work?

Any client that speaks MCP over Streamable HTTP: Claude (web, desktop and Claude Code), ChatGPT in developer mode, Cursor, Windsurf, Zed and others. There is nothing to install and no key to paste.

Do I need an API key or an account?

No. The connector is open and read-only. Add the endpoint URL to your MCP client and start asking questions.

What is never returned?

Anything not editorially approved and published: pending funding rounds and pending companies or people awaiting review, draft articles, and archived records. If it is not live on the site, the connector cannot see it.

500 calls per connection per day (per IP address for clients that don’t keep a session), a short burst limit of 60 calls a minute, and at most 200 rows per call.

What happens at the limit?

The call comes back as a tool error naming the limit and the time it resets, together with a pointer to the bulk dataset files. Nothing breaks: the connection keeps working and the allowance starts again at midnight UTC.

The tool name, and a salted irreversible hash of the caller so the daily limit can be applied. Those counters are deleted after 48 hours. We never store the arguments, the results or a readable address.

New records arrive twice a day; the dataset files are rebuilt every hour. Anything approved by the editor is available to the connector immediately afterwards.

Are amounts converted into one currency?

Never. Every amount is reported in the currency it was announced in, so totals are given per currency.

Can I reuse the results?

Yes, under CC BY 4.0: attribute Builders in Fintech and link back to the canonical URL that every record carries.

Read Builders in Fintech from your assistant

Every company, investor, person, funding round, event and published story we hold, live over the Model Context Protocol. Every record is reviewed by the editor before it is published, and carries a link to its source and its canonical URL so your assistant can cite it. The data is CC BY 4.0. No key, no signup.

https://buildersinfintech.ai/mcp

One line in Claude Code

claude mcp add --transport http builders-in-fintech https://buildersinfintech.ai/mcp

Other clients take the same URL — see Connect in 3 steps below.

Call it from a terminal

curl -s https://buildersinfintech.ai/mcp \
  -H 'content-type: application/json' \
  -H 'accept: application/json, text/event-stream' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"list_funding_rounds","arguments":{"country":"gb","limit":5}}}'

The reply lists five recent rounds from the United Kingdom, and carries a usage block telling you how much of your daily allowance is left.

What the assistant can call

Tool	What it does	Access
`search_organizations`	Search fintech companies and investors by name, kind or country. Returns public profiles with canonical URLs. Arguments: q (or query): text to search for, matched against the name. With no q the tool returns a plain alphabetical list (first 25 by default). Unknown arguments are rejected.	Open — no key Counts toward daily limit
`get_organization`	Full public profile for one organization by slug: description, country, funding rounds (raised or led), people, any disclosure, our published coverage of it (articles, newsletter issues, podcast episodes), and its products (company-provided, not part of the CC BY dataset).	Open — no key Counts toward daily limit
`search_people`	Search founders, operators and investors covered by the publication. Arguments: q (or query): text to search for, matched against the full name. With no q the tool returns a plain alphabetical list (first 25 by default). Unknown arguments are rejected.	Open — no key Counts toward daily limit
`get_person`	Full public profile for one person by slug: bio, roles and coverage.	Open — no key Counts toward daily limit
`list_funding_rounds`	Approved funding rounds, newest first, with optional filters and keyset pagination. Pass the returned next_cursor fields (after_date, after_id) to get the next page.	Open — no key Counts toward daily limit
`largest_rounds`	Largest disclosed equity rounds, ranked per currency (amounts are never converted between currencies).	Open — no key Counts toward daily limit
`most_active_investors`	Investors ranked by the number of approved rounds they appear in. Category defaults to equity. With a country, only rounds of companies headquartered in that country are counted; the country may be a slug ('italy') or a two-letter ISO code ('IT'), case-insensitive.	Open — no key Counts toward daily limit
`funding_summary`	Counts for a slice of the funding database: rounds, countries, totals per announced currency, undisclosed rounds and rounds per category. Note: rounds, per_currency and undisclosed count equity rounds only, while by_category covers every category (equity, debt, fund, grant, ipo), so the category split can sum higher than the round count.	Open — no key Counts toward daily limit
`get_funding_index`	The weekly Builders Fintech Funding Index of approved equity rounds (ISO weeks). Arguments: week (optional, 'YYYY-Www', e.g. '2026-W39') returns that week; otherwise the latest weeks, newest first. weeks (optional, 1–104, default 1) sets how many. Count Index = rounds this week / average of the prior 12 weeks × 100. USD median index uses disclosed USD rounds only (needs 5). Amounts are never converted. Published values are frozen; revised_* shows later corrections. coverage_affected is true when a coverage change falls in the week or its 12-week baseline; coverage_note then explains how the index overstates activity. Unknown arguments are rejected.	Open — no key Counts toward daily limit
`get_investor_scorecard`	Follow-on scorecard for one investor by slug: of the portfolio companies it first backed at least the window ago, the share that raised a later approved round (equity or IPO) within the window, at least gap_days after entry. Only returned when the investor has at least min_eligible eligible companies. Arguments: slug (required). Unknown arguments are rejected.	Open — no key Counts toward daily limit
`search_products`	Search fintech products by text, category, market, API availability or customer type. Company-provided, not part of the CC BY dataset. Each result carries provided_by, last_confirmed_at and sponsored; sponsorship never changes ordering. Arguments: q (or query) matched against name and short description; with no q the tool lists products alphabetically. cursor is the opaque next_cursor from a previous page. Unknown arguments are rejected.	Open — no key Counts toward daily limit
`get_product`	Full public listing for one product by slug: description, category, markets, customer type, API, pricing model, links and the company. Company-provided, not part of the CC BY dataset. Each result carries provided_by, last_confirmed_at and sponsored; sponsorship never changes ordering.	Open — no key Counts toward daily limit
`search_content`	Full-text search across published news, newsletter issues, podcast episodes and public entity pages. Arguments: q (or query): text to search for, at least 2 characters — required, this tool has no plain-list mode. Unknown arguments are rejected.	Open — no key Counts toward daily limit
`get_content`	One published item by type and slug, with its body text (truncated), topics, countries, organizations and people. News items also include a corrections array (date + note), newest first.	Open — no key Counts toward daily limit
`search_podcast_knowledge`	Search attributed facts and question-and-answer pairs from Builders in Fintech podcast episodes (as stated by the guest, not independently verified). Arguments: q (or query) matched against fact text, and against questions and answers; organization (slug) = facts linked to that company or investor; person (slug) = facts said by that person; fact_type (figure, fact, plan, view); episode (podcast episode slug); limit; cursor (next_cursor from the previous page). With no q the tool lists the newest facts. The qa array is returned only when q is given and honours q and episode only. Unknown arguments are rejected.	Open — no key Counts toward daily limit
`list_events`	Approved fintech events, upcoming by default, with optional country, format, topic and month filters.	Open — no key Counts toward daily limit
`list_topics`	Every public topic with its slug and URL, for use as a filter in the other tools.	Open — no key Counts toward daily limit
`list_countries`	Countries that have public coverage, with company and content counts.	Open — no key Counts toward daily limit
`about`	What this database covers, how to cite it, and where the bulk downloads live.	Open — no key Counts toward daily limit

Search tools take the search term as q (query is accepted as an alias). Unknown arguments are rejected with an error that lists the valid ones, so a client can correct its call.

Four questions to start with

Journalist

This week in fintech funding

Using the Builders in Fintech connector: list the fintech funding rounds in Saudi Arabia announced since last Monday with list_funding_rounds. If fewer than three rounds come back, widen the window to the last 30 days and tell me you did. For each round give the company, amount and currency exactly as reported — never convert between currencies — the round type, the investors and the announcement date. Then call get_organization for the two largest companies and add one line each on what they do. Cite each record's url (and page_url where one is given).

Investor

Company briefing with coverage

Using the Builders in Fintech connector: call search_organizations for "Circle", then get_organization for the best match. Give me a briefing with: what the company does, headquarters and country, its funding rounds in chronological order with amounts in their original currency, its known people and their roles, and the articles, newsletter issues or podcast episodes in the coverage list. If the coverage list is empty, say so rather than guessing. Link the company and every cited item to its url.

Analyst

Largest rounds this year plus totals

Using the Builders in Fintech connector: show me the largest fintech equity rounds announced this year with largest_rounds (year = 2026), grouped by currency — never convert amounts between currencies. Then call funding_summary from 1 January 2026 and give me the totals: number of rounds, totals per currency, undisclosed rounds and the split by category. Read the result's notes field and repeat the caveat it gives about which counts are equity-only. Link each company to its Builders in Fintech page and cite each record's url (and page_url where one is given).

Researcher

What we have written on a theme

Using the Builders in Fintech connector: call search_content for "stablecoins" with kinds ['newsletter','podcast'] to find what we have published on that theme. Then call get_content on the two most relevant items and summarise each in a short paragraph, naming the companies and people they discuss. Give me the newsletter issue number or the podcast episode number where the record carries one, and cite each record's url. If nothing comes back, say so rather than answering from general knowledge.

Clients that support MCP prompts will find all four ready to run under the connector’s name.

Connect in 3 steps

Claude

Settings → Connectors → Add custom connector.
Name it Builders in Fintech and paste https://buildersinfintech.ai/mcp.
Set the tool permissions to “Always allow”.

ChatGPT

Settings → Apps & Connectors → Advanced → turn Developer mode on.
Create, paste https://buildersinfintech.ai/mcp, and set Authentication to “No authentication”.
Enable it in the tools menu of the chat where you want to use it.

Cursor

Add it to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "builders-in-fintech": {
      "type": "http",
      "url": "https://buildersinfintech.ai/mcp"
    }
  }
}

Any other client: choose Streamable HTTP, paste the URL, no authentication.

Menu paths as of September 2026 — clients rename these often.

Limits, licence and privacy

500 calls per connection per day (per IP address for clients that don’t keep a session), plus a short burst limit of 60 calls a minute.
At most 200 rows per call. For bulk analysis, use the open dataset files instead.
Only approved, published records are returned. Amounts stay in the currency they were announced in and are never converted.
Results are CC BY 4.0: attribute Builders in Fintech and link to the canonical url carried by every record.
We count calls per tool and keep a salted, irreversible hash of the caller so the daily limit can be applied. No arguments, no results and no readable addresses are stored.

Connector manifest · Open dataset files · Instructions for agents · Privacy

Frequently asked

What is MCP?: The Model Context Protocol is an open standard that lets an AI assistant call a data source directly instead of guessing from what it read during training. You add one URL to your assistant and it can then query our database while it answers you.
Which assistants work?: Any client that speaks MCP over Streamable HTTP: Claude (web, desktop and Claude Code), ChatGPT in developer mode, Cursor, Windsurf, Zed and others. There is nothing to install and no key to paste.
Do I need an API key or an account?: No. The connector is open and read-only. Add the endpoint URL to your MCP client and start asking questions.
What is never returned?: Anything not editorially approved and published: pending funding rounds and pending companies or people awaiting review, draft articles, and archived records. If it is not live on the site, the connector cannot see it.
What are the limits?: 500 calls per connection per day (per IP address for clients that don’t keep a session), a short burst limit of 60 calls a minute, and at most 200 rows per call.
What happens at the limit?: The call comes back as a tool error naming the limit and the time it resets, together with a pointer to the bulk dataset files. Nothing breaks: the connection keeps working and the allowance starts again at midnight UTC.
What do you log?: The tool name, and a salted irreversible hash of the caller so the daily limit can be applied. Those counters are deleted after 48 hours. We never store the arguments, the results or a readable address.
How fresh is it?: New records arrive twice a day; the dataset files are rebuilt every hour. Anything approved by the editor is available to the connector immediately afterwards.
Are amounts converted into one currency?: Never. Every amount is reported in the currency it was announced in, so totals are given per currency.
Can I reuse the results?: Yes, under CC BY 4.0: attribute Builders in Fintech and link back to the canonical URL that every record carries.