What's New

Recent updates to GrantLoop.

August 2026

Meet Louellen 🎉

Your agent has a name. She's Louellen, and you'll see her by name throughout the app from now on — including when you start a new chat. Same agent, same work; she's just someone you're working with rather than a panel on the side of the screen.

She also has her own email address now. When a scheduled task finishes, or she needs a decision from you, she can write to you — and you can just hit reply. Your answer goes straight back into the work: no logging in, no hunting for the right chat. Attach a file to your reply and she reads it. Emails show up in your chat history alongside everything else, so the thread reads as one conversation whether it happened in the app or in your inbox.

Every email from Louellen is clearly labelled as coming from an AI. She'll never pretend otherwise.

The best models available, for fewer credits

GrantLoop now runs on the newest generation of OpenAI models. There are two, and you pick between them under the message box:

GPT-5.6 Terra — the default. It handles almost everything.
GPT-5.6 Sol — the stronger of the two, for the genuinely hard calls. It spends credits about 2.5× faster, so save it for where the extra judgement earns its keep.

At the same time we've spent a lot of this month on benchmarking and internal research, finding places where GrantLoop was spending more than it needed to without getting anything back for it. Recurring tasks were the biggest one — they now run at up to 50% fewer credits, across the whole run — and there were plenty of smaller ones alongside it.

None of it makes Louellen less capable. All of it goes to you: most organizations should get 20–50% more out of their monthly credits from here on.

Folders

You can organize your workspace properly now. Group sheets, documents and chats into folders in the sidebar, nest them, and file your Reference Materials the same way. Select several things at once to move them together.

Folders live with your account rather than your browser, so they follow you between devices and look the same for everyone on your team. And you can just ask Louellen to do it — make a folder, file things into it, rename a sheet, tidy up.

A simpler interface

We've stripped a lot out. The app opens in one clean step instead of four stacked loading screens, Settings has been rebuilt, and the chat box no longer lags on long conversations.

The thread through all of it: the point of GrantLoop is working with Louellen, and everything else should get out of the way of that.

Autopilot does more, and spends less

Autopilot is more thorough than it was. It checks you're actually eligible for an opportunity before doing the expensive research on it, keeps your files and folders organized as it works, and shows you what it moved instead of quietly rearranging things. It also gets through all of that on noticeably fewer credits.

Also

Dangerous but Awesome Mode is now on by default. Louellen takes actions on your behalf without stopping to ask at every step. You can turn it off in Settings → General.
Better memory. Louellen keeps a clearer record of what she's done and what you've told her, so she picks up where you left off more reliably.
Bug fixes all around.

A note on legacy plans

A small number of long-time customers are on early arrangements we set up before GrantLoop had usage limits at all. Those accounts are moving onto monthly limits that are deliberately generous — set well above how those organizations actually use the product, so ordinary work shouldn't come close to them.

We want to be straightforward about why. GrantLoop has become a great deal more capable over the past year, and the models behind that capability cost real money to run. Leaving a handful of accounts genuinely uncapped isn't something we can carry sustainably as the platform grows — and we'd much rather set a fair, workable limit than quietly make the product cheaper and worse for everyone.

July 6, 2026

Views are here 🎉

This is the biggest update we've shipped in a while, and we're really excited about it: your prospecting and tracking tables now have views — including a board view that shows your pipeline as cards you can drag from one stage to the next.

Grant work is pipeline work. A spreadsheet tells you what's in the pipeline; a board shows you where everything stands — what's in research, what's drafting, what's waiting on a funder. Now you get both, from the same table:

Board view — cards grouped by status (or any dropdown field). Drag a card to move it forward, click it to edit, add new ones right where they belong.
As many views as you like — a table sorted by deadline, a board of this quarter's applications. Each view keeps its own sort and filters, and they're shared with your whole team.
Totals at a glance — see counts or dollar totals for each group, right on the board.

Your agent knows about views too — just ask it to set up a board or reorganize your groups.

Autopilot works the board

Autopilot now leans on this. It keeps a Dashboard board of your pipeline, lists To-Dos on every opportunity so you always know the next step, and tidies up after itself — clearing out past-deadline items and keeping statuses current. Its weekly emails link you straight to the Dashboard.

Budgets and reports, in your format

Files the agent creates for you — budgets, reports, summaries — come out noticeably more polished now. Even better, they can follow your formats: share the budget template or report layout your organization uses and ask the agent to remember it going forward. It'll save the format to your Reference Materials and follow it every time after.

Twice the grant history

We've doubled the size of our historical grants dataset. The agent now searches over 15 million historical grant records from foundation tax filings — up from around 10 million. Coverage used to get thin before 2018; it now extends all the way back to 2009, and includes far more grants from this past year, so the agent sees funders' most recent giving, not just where they were a few years ago.

In practice this means fuller funder giving histories, sharper "how much should we ask for" estimates, and more matches when you ask who funds organizations like yours.

Also new

New field types — dates (with a real picker, plus "Rolling" and "TBD" deadlines) and number formats like currency and percent.
Change fields freely — rename a field or change its type, and your data comes along.
Invite your team — Settings → Members to invite teammates and manage roles. Every plan includes up to 5 seats.
Knowledge is now Memory — same feature, clearer name.
The agent edits your real files — hand it a Word doc from your Reference Materials and it works on the actual file, formatting preserved.
Emails with real attachments — ask the agent to email you a file it made and it arrives as an actual attachment, not just a link.
Focus mode — tuck away the sidebar and chat for distraction-free writing.
Lots of polish — friendlier dates, multi-line cells, smoother scrolling, and sturdier long-running conversations.

As always, hit us up with feedback — this one's been months in the making and we can't wait to see your pipelines on it.

June 23, 2026

Best Contact: Find the Right Person to Reach Out To (Beta)

The strongest grant relationships rarely start with the program officer everyone else emails. They start with a real connection — someone on a funder's board or staff who already cares about your work, or who shares a thread with someone on your team.

Now you can ask your agent to find that person. Just tell it who you're targeting:

"Who's a good person to reach out to at the Cascade Community Foundation?"

The agent researches the funder's entire board, staff, and officers — not just the obvious contact — and digs into who these people actually are: their hobbies, the organizations they volunteer with, where they work, and where they went to school. It's looking for two kinds of connection:

They care about your work — personally, professionally, or academically. A trustee whose side project is writing a cookbook is a real opening for a soup kitchen; a program officer who built their career in youth literacy is far more than a job-title match.
They're connected to your team — someone who went to the same school as your director, shares a hobby with a board member, worked at an organization you partner with, or sits on a board alongside one of your people.

You get back a short, ranked list: who to reach out to, why they're a fit, how to reach them, and — when an email is available — a drafted intro ready to send. The obvious official contact is still there too, clearly labeled as the door you already have.

This feature is in beta and we're actively refining it. Finding the right relationship is one of the most important — and most overlooked — parts of winning a grant, so we're investing heavily in getting this right. Your feedback makes it better.

Try it: "Find me a good contact at X funder."

May 14, 2026

A note on recent reliability

Some of you have noted intermittent issues over the past two days — conversations dropping or stalling during complex tasks. We suspect this is throttling from one of the partner AI companies (OpenAI, Anthropic, Google) we rely on for our infrastructure, as those services scale to meet broadly increased general demand for AI.

We've taken measures to diagnose the immediate issue and route our requests through a path we expect to receive higher priority. In preliminary testing, this has resolved the issues we were seeing.

If you continue to experience these issues after today, please message support and we'll help you sort it out.

Sincerely, Ben

April 20, 2026

GrantLoop Autopilot (Beta)

Imagine opening your inbox every week to a fully written grant application, ready to review and submit. That's GrantLoop Autopilot. We'll be refining this in the coming weeks — your feedback helps us make it better.

Turn it on, and your AI agent takes over the entire grant pipeline — from finding opportunities that fit your organization, to pulling application questions from funder portals, to writing complete drafts using your materials and past applications. It tracks everything in a prospecting sheet and a tracking sheet that stay updated automatically. When it needs your input, it asks. When it doesn't, it just keeps going.

All you have to do is review and submit. And if you give the agent portal access, it can even handle that part too.

GrantLoop Autopilot is free and included with every plan.

How to get started:

Just tell your agent: "Start managing my grants on autopilot."

Or go to Recurring Tasks in the sidebar and toggle the Grant Pipeline Manager task to on.

March 24, 2026

AI-Generated Excel Budgets

The agent can now create real Excel spreadsheets with working formulas — not just text tables. Ask it to build a grant budget and it generates a downloadable .xlsx file with a funder-facing Budget tab and an Inputs tab with editable assumptions.

It learns your organization's budget style by searching through your Reference Materials and connected integrations. The more context it has — past budgets, indirect cost rates, how you categorize line items — the closer the output matches how you actually do things.

"Make me a $50,000 budget for a community gardens program with matching funds."

See the full guide: Budgets

Web Browsing & Automated Form Filling

The AI can now click buttons and type on websites. Previously, the agent could search the web and read pages — but it couldn't actually interact with them. Now it can navigate complex funder portals, retrieve application questions, and fill out forms with your grant responses.

Give it a portal URL and it will click through pages, extract what it finds, and bring the information back into GrantLoop. If you have a completed application, the agent can go to the funder's portal and prefill the form for you.

A note on responsibility: This is a new and powerful capability. We recommend reviewing the agent's browser actions as they happen and double-checking any submissions before they go out. As with any AI tool, you're ultimately responsible for what gets submitted on your behalf.

See Web Browsing for details.

Voice Input

You can now dictate messages to your AI agent instead of typing. Hold the microphone button in the chat panel to speak — your words appear in real time as you talk. Release to stop recording. Voice input works everywhere the chat panel appears.

Manage Recurring Tasks from Chat

You no longer need to navigate to the Recurring Tasks page to set up scheduled agents. Just tell the agent in chat: "Set up a weekly task to check for new grants every Monday at 9am." The agent can create, update, pause, resume, and delete recurring tasks directly from the conversation.

Persistent Edit Proposals (Beta)

When the AI proposes changes to your grant document, those edits are now saved automatically. If you navigate away or close the tab, your pending edits will still be there when you come back — no more lost proposals.

This feature is in beta. If a large edit doesn't appear right away, try refreshing the page or navigating away from the document and then back to it. We're actively working through these issues.

Improved Reliability

Faster stream recovery when connections drop, better handling of long-running workflows, and fixes for recurring task scheduling.

March 15, 2026

A Smarter Agent

We've upgraded to GPT-5.4 — the most capable AI model available today. The agent should feel noticeably more intelligent: better drafts, reduced lag, more reliable output, and the ability to handle longer, more complex workflows without breaking down.

We've also removed shortcuts like "AI Quick Start" and "AI First Draft." You don't need them anymore. Just talk to the agent directly — tell it what you need, and it'll figure out the rest. Create a new project and start chatting.

The grantwriting workflow now uses a multi-agent approach behind the scenes: specialized sub-agents handle research, writing, and assembly in parallel, producing higher-quality drafts in less time.

We encourage you to push the boundaries of what the agent can do. It's far more capable than it was even a few weeks ago — try giving it ambitious, multi-step tasks and see what happens.

March 13, 2026 — A Note on Recent Issues

Over the past few weeks, many of you have experienced slow responses, crashes, and formatting problems with generated text. We want to be transparent about what happened and sincerely apologize.

Two things happened at the same time:

Rapid growth in usage. More users and heavier workloads put strain on our infrastructure, leading to slow responses, dropped connections, and intermittent outages.
Increased model capabilities. The AI model became significantly more capable, which meant it was doing more work autonomously for longer stretches — demanding more resources from our systems than we were prepared for.

Combined, these caused a rough experience for many of you — and that's not the standard we hold ourselves to. We strive to make GrantLoop feel magical, and this hasn't been that.

If you were affected, please contact support. We'd like to offer a refund and find a way to make it up to you.

We've shipped fixes to reduce timeout errors and increase speed, resolved formatting bugs in generated drafts, and improved overall draft quality. Things should feel noticeably better going forward.

Thank you for your patience. We're grateful you're here, and we're committed to earning your trust.

Gratefully, Ben Yao, CEO & Founder

March 2026

Integrations

Connect your existing tools so the AI agent can work with the data you already have. Currently supported integrations include Google Drive, Google Docs, Google Sheets, SharePoint, Airtable, Fathom, and Fireflies.

To get started, visit the Integrations page from the sidebar, or just ask the agent in chat — it'll walk you through connecting.

Coming soon: We're working on giving the agent the ability to take action within your connected tools — things like data entry in Airtable, file management in Google Drive, and admin tasks on your behalf. We're testing for safety and reliability before rolling this out.

Knowledge (now called Memory)

We're introducing Knowledge — a way to give your AI agent standing instructions, just like you would with a fellow coworker.

You can access it from the sidebar (now labeled Memory) to add entries, or just tell the agent in chat and it'll remember for next time.

Examples of things you might tell your agent:

"Whenever I start a new draft, look through my Google Drive for our historical grant applications and use them as a style guide."
"Check our Airtable to see if we've applied to or won grants from a funder before."
"If we ever need board composition details or specific financial documents, they're in our SharePoint site — search there and insert links into the grant."
"Our former programs officer Becky retired — don't mention her in any applications going forward."

Task Agents

The AI agent can now delegate work to independent task agents that run at the same time. When the agent needs to do multiple things — like research a funder while pulling data from your prospecting table — it spins up task agents to handle each piece concurrently, so you get results faster.

Grantwriting

Improved reliability when the agent makes edits and revisions to your grant documents.

January 2026

Unified Chat Panel

The AI chat interface has been completely redesigned. Previously, prospecting and grantwriting had separate chat experiences. Now there's a single unified chat panel that works across all project types.

What this means for you:

One consistent interface everywhere
The AI automatically knows what project you're working on
Seamlessly switch between prospecting and grantwriting tasks in the same conversation

Recurring Tasks

Set up AI tasks to run automatically on a schedule.

Use cases:

Weekly prospecting research for new funding opportunities
Regular deadline monitoring and status updates
Automated enrichment of your prospecting tables

How it works:

Create a recurring task from the sidebar
Give it a task description and schedule (daily, weekly, etc.)
The task runs in the background and you can review results anytime

CSV Export

Export your prospecting tables to CSV with one click. Find the export button in the toolbar above your prospecting table.

Cleaner sign up flow with better error handling
Dedicated password reset page
Existing account detection during sign up (redirects you to login instead)