GrantLedger in ChatGPT and Claude
Ask an assistant which calls fit a startup in Portugal, a musician in the UK or a research lab in Austria, what a loan guarantee in Germany requires, or what to apply to first from your own screening. It reads the same ledger as this site and links every answer to the official publisher page.
What the plugin does
- Finds calls. It searches every current call in the ledger by country, type, audience and words, and links each one to the official page.
- Explains a call. What it requires, which document proves each condition, what the jargon means for you.
- Screens a project. Describe a company or a project; it runs the same rule-based screening as this site and shows the verdict and the rule for each call.
- Saves a profile and sets the weekly alert. Only after it shows you the profile and you say yes. New calls that fit are then emailed every Monday.
- Helps you plan, write and score an application. From your profile, the call and the criteria on the official page. You submit yourself.
What it costs
The plugin is part of Pro: one payment of 9 euros (Solo, 1 saved profile) or 19 euros (Team, 3 saved profiles), not a subscription. There are no credits and no fee per question, per draft or per application. The explaining, writing and scoring run on the assistant you already use, under your own ChatGPT or Claude plan; GrantLedger supplies the calls, the rules and your profile. Searching the ledger on this site stays free without Pro.
Past the jargon, without an intermediary
Finding a call is half of the work; reading it is the other half. Once the ledger is connected, the assistant explains in plain words what a call requires, which document proves each condition and what a term such as "de minimis" or "consortium" means for you. You stay in charge: you apply yourself, with the funder, and nobody takes a share of the funding.
One address, then sign in
GrantLedger is a Model Context Protocol (MCP) server. Add this address as a connector; the assistant opens a GrantLedger page where you sign in with your Pro email (a code by email, no password) and click Allow. No key to copy.
https://grantledger.eu/mcpGet Pro One payment, no subscription. No account yet? The sign-in page offers Pro too, and brings you back once paid.
Tools: search_grants, get_grant, get_fit, my_profiles, plus search and fetch for ChatGPT, all read only, and screen_profile and save_profile. Sixty requests per minute.
ChatGPT
- Connect the plugin. ChatGPT opens the GrantLedger sign-in page.
- Type your Pro email, then the six-digit code we email you, and click Allow. No Pro account yet? The same page offers it and brings you back once paid.
- In a chat, enable GrantLedger and ask. Deep research also uses it through search and fetch.
Workspace without the plugin directory: Settings, Security and login, turn on Developer mode, then Plugins, Add, Create MCP App with the address above and OAuth.
Claude
- Settings, Connectors, Add custom connector. Name GrantLedger, paste the address, Connect, sign in.
- Claude Code, in a terminal, then /mcp to sign in:
claude mcp add --transport http grantledger https://grantledger.eu/mcpCursor, Codex, Windsurf, Zed
Same address as a streamable HTTP server; each opens the sign-in page on first use.
Your screening in the chat
Screen your company on this site, then paste the result link in the chat. The link looks like this:
https://grantledger.eu/fit/<id>The assistant calls get_fit and receives the profile you confirmed and every call screened for it with the rule each verdict relies on. Then ask it to rank, explain a gap, draft a first email to the funder, or compare two calls.
Once you have saved a profile, no link is needed: ask "what is new for my profile?" and the assistant reads your saved profiles and the calls the Radar emailed you through my_profiles.
Screen and save from the chat
Screen any project. Describe your company or a project and ask: "Screen this project against the ledger." The assistant asks what it does not know and runs the same screening as this site through screen_profile: every call with its verdict and the rule behind it. Screen as many projects as you like, a second product, a what-if, a client; the limit is twenty screenings a day. A screening is a snapshot of that day's ledger: ask again later for a fresh one.
Save the ones to follow. Ask: "Save this profile and alert me about new calls." The assistant shows you the complete profile and saves it through save_profile only after your yes. A saved profile is screened again every Monday and new calls that fit are emailed to you, as if you had saved it on this site. Your plan sets how many profiles you can save; screening has no such limit.
Connected before this existed? That connection stays read only: disconnect GrantLedger in your assistant and connect it again. The personal link below neither screens nor saves. Change or remove a profile on your page.
Plan, write and score an application
The assistant you already pay for does the writing; the ledger gives it your profile, the call and the readiness gaps. Pick a call from your screening and paste one of these:
- "Plan my application to this call from the readiness gaps in my screening, counted back from the deadline. Say if the deadline is realistic."
- "Write a first draft of my application to this call. Read the form and the criteria on the official page, show me the outline first, and put a placeholder where you miss a fact."
- "Score my draft against the evaluation criteria published for this call, criterion by criterion, and give me the three changes that gain the most points."
- "This call needs a consortium. Show me where consortia look for partners for it, and draft my partner offer from my profile."
The assistant reads the form and the criteria on the publisher's page; if it cannot open them, paste or upload the call text. A score is the assistant's reading of the published criteria, not the funder's evaluation and not a verdict of this site. Your draft stays in the chat: no tool takes it. Check the funder's rules on AI tools in applications. Prefer a service that writes and submits with you? See how that compares. New to public funding? Start with the guides.
Tools without OAuth
A script, a JSON client or an older tool that cannot sign in takes your personal link instead. It carries your key, so keep it to yourself.
Sign in to see your personal link.
The plain JSON API is described in openapi.json (OpenAPI 3.1) and takes the key or an OAuth token as a bearer: /api/grants.json, /api/grant.json and /fit/{id}.json.
What the assistant can and cannot do
- Verdicts come from fixed rules on this site, never from the assistant. It reads them; it does not decide them.
- The ledger holds what official sources publish. If a call is not in the results, it is not in the ledger; ask the assistant to say so rather than guess.
- An assistant reads, screens what you describe, and saves a profile only when you ask and confirm it. It cannot change or remove a saved profile, your account or your payment.
- The connector sends nothing to us except the queries, the ids it looks up and the profile you ask it to screen or save. See the privacy notice.
Questions that work well
- "Which open calls in Spain fund industrial SMEs buying machines? Link the official pages."
- "EU-wide programmes for a deep-tech startup under three years old, deadline after next month."
- "Open calls for an independent musician in the UK. Which ones accept an individual?"
- "Research funding in Austria for a postdoc in life sciences. Explain the eligibility rules in plain words."
- "Which three calls should I apply to first and why? My screening: https://grantledger.eu/fit/<id>"
- "For the second call, what does 'to check' mean and what document proves it?"