Features
Everything, by stage
Seven stages. Each one is a set of concrete jobs, not a slogan.
01 · Discover
Continuous discovery from the original sources
A scheduled job runs every 6 hours against the official APIs and feeds. Opportunities are deduplicated by funder and opportunity number, so one notice appears once no matter how many places publish it.
- Simpler.Grants.gov and Grants.gov for federal notices, SBIR.gov for open small business topics, the NIH Guide feed for notices and reissues, SAM.gov for contracts and BAAs.
- HTML only sources such as CDMRP, NSF Seed Fund, Wellcome Leap, and the Michael J. Fox Foundation are scraped on a schedule, so foundation and DoD programs sit in the same table as federal grants.
- Deadline changes, reissues, and expirations are detected on every run and the affected matches are rescored, so a queue never carries a dead opportunity.
- Manual entry for the opportunity a program officer emailed you, with the same scoring and tracking as everything else.
02 · Qualify
Hard eligibility first, then fit
Every new or changed opportunity is scored against your company profile with a deterministic rubric and an AI written rationale. Fit score from 0 to 100, win probability anchored to the published base rate, expected value, and a plain language verdict.
- Eligibility flags for entity type, size, ownership, work location, the 67 percent Phase I and 50 percent Phase II work share rules, and foreign components.
- The rubric weighs programmatic fit, funder history with similar work, base rate for that institute and phase, technical readiness versus stage requested, team completeness, relationship, and time to deadline.
- Strings attached are extracted from the notice: data rights, reporting cadence, cost sharing, audits, publication and data sharing, foreign disclosure.
- A deliverables summary tells you what winning would obligate you to produce before you apply.
03 · Decide
One click approval with the numbers in front of you
Matches above your strictness threshold land in a queue. A daily digest emails the new ones. Each card shows fit, win probability, award ceiling, expected value, and the deadline, so a go or no go decision takes a minute, not a meeting.
- A strictness slider sets the fit threshold and previews the count of queued opportunities, the sum of award ceilings, and the likely grant volume at that setting.
- Approve creates the application. Decline records the reason, which feeds the next scoring calibration.
- Funder history for your team is on the card: previous applications, outcomes, and total awarded.
04 · Apply
AI drafted applications that never miss a deadline
Approval downloads the forms, converts the notice into a section checklist, and drafts every section from your verified company fact library. Anything missing raises an information request assigned to the person with the right role.
- Sections are drafted from facts, not invented. Each fact has an owner and a last verified date.
- PDF forms are filled from sections and facts. Word and page limits are tracked per section.
- An AI review panel critiques the draft against the funder's own criteria before you submit.
- Submission by portal automation, email, or generated PDF, always after explicit confirmation. Files are mirrored to your Dropbox under Grants, Funder, Opportunity.
- Reminders at 30, 14, 7, 3, and 1 days before every deadline, in email and in your task manager.
05 · Monitor
Inbound mail classified against open applications
A monitored mailbox is scanned every 30 minutes. Receipts, requests for just in time information, summary statements, award letters, and rejections are classified and attached to the right application, and status updates without anyone forwarding an email.
- Expected decision dates on a timeline next to deadlines and award periods.
- Tasks sync both ways with Melina or Todoist every 15 minutes.
- Every interaction with a funder is kept, so relationship memory survives staff changes.
06 · Deliver
Deliverable tracking after the award
An award letter becomes a compliance calendar. Terms are extracted, deliverables are created with due dates, and the same reminder cadence applies to progress reports, financial reports, invention statements, and closeout.
- Deliverable content is generated from the Dropbox company folder and the fact library, with information requests for anything missing.
- Spend versus budget from PolyAccounts, with exceptions flagged before a report is due.
- Success fee invoices are raised through Stripe only when an award is recorded.
07 · Learn
Learns from every outcome with TALR
Every application runs a Think, Act, Learn, Repeat cycle. Outcomes, reviewer scores, and summary statement critiques update the assumptions and the funder model, which feed the next Think step and recalibrate the scoring.
- Predicted win probability is plotted against actual outcomes so the multiplier curve can be recalibrated.
- Critiques raised independently by two or more reviewers are tagged and carried into the resubmission plan.
- Weekly strategy notes summarize where the pipeline is strong, where it is thin, and what to change.
Sources
Official APIs first, scraping only where there is no API
Simpler.Grants.gov
API Federal notices with structured search and daily extracts
Grants.gov search2
API Federal notices, forms, and the system to system channel
SBIR.gov
API Open SBIR and STTR solicitations and topics across agencies
NIH Guide
Feed Notices, reissues, and expirations the day they post
SAM.gov
API Contracts, BAAs, ARPA-H, BARDA, and NIH contract solicitations
CDMRP
Scraped DoD medical research programs and their annual topic lists
NSF Seed Fund
Scraped NSF SBIR and STTR solicitations and Project Pitch windows
Foundations
Scraped Wellcome Leap, the Michael J. Fox Foundation, and more on request
Each source can be enabled per team, with its own credentials where the source requires them. Manual entry covers the opportunity a program officer sent you directly. Every opportunity is deduplicated by funder and opportunity number, so a notice published on three portals appears once.
Scoring
What is on every match card
Fit score
0 to 100 from an eight factor weighted rubric with written anchors
Win probability
Base rate times a fit multiplier, adjusted for resubmission and prior awards, clamped between 1 and 75 percent
Expected value
Probability times usable award, less the cost to apply, less probability times the cost of compliance
Eligibility flags
Entity type, size, ownership, place of business, work share, foreign components, registrations
Strings attached
Data rights, reporting cadence, audits, cost share, publication and data sharing, foreign disclosure
Deliverables summary
What winning would obligate you to produce, and when
Funder history
Your team's applications, outcomes, and total awarded with this funder
Verdict
Go, conditional, or no go, with the rationale in plain language
The rubric is published in Expected value math for grant applications.
Compared
What most grant tools do, and what epiensos does instead
| Typical | epiensos |
| Fit score as a percentage | Fit score anchored to the published base rate for that institute and phase, then converted to a win probability and an expected value |
| Weekly email of new grants | Discovery every 6 hours, rescoring on every deadline change, and a queue that only holds live opportunities |
| AI writes a draft | AI drafts from a verified fact library, raises information requests for what it cannot source, and an AI panel reviews the draft against the funder's own criteria |
| Deadline calendar | Reminders at 30, 14, 7, 3, and 1 days in your task manager, and submission only after explicit confirmation |
| Post award spreadsheet | Award terms extracted into deliverables with generated content, spend versus budget from your accounting system, and automated submission where supported |
| Win rate chart | A TALR cycle on every application, critiques tagged by criterion, predicted versus actual calibration, and weekly strategy notes |
Integrations
Works with your tools
Dropbox
System of record for every application and deliverable file
Todoist
Two way task sync for deadlines and information requests
Melina
Internal task driver with the same two way sync
PolyAccounts
Spend versus budget on every award, signed integration events
Gmail
Monitored mailbox classified against open applications
Stripe
Monthly subscription and success fee invoices
Plus a REST API with an OpenAPI document, an MCP server, per team API keys, and an agent driver. See For agents.