The system of record for how work actually happens.
Fyboard is a records-first operating system for modern execution. We build engines, not apps. Permissions and Audit sit at the bottom — four-tier organisation RBAC resolved on every request, hierarchical row scoping with time-bounded grants, and mutation auditing with before/after diffs. FyBrain (intelligence), Workflow (orchestration) and FyDrive (files and documents) sit on top of those, and every module sits on top of all of it.
Fyboard is built inside the digital arm of an Indian EPC group with ₹1,300 crore+ in FY25-26 revenue. The platform was forged on a real construction site — not in a vacuum, not in a deck, not in a venture office. The schema graph behind it is roughly 1,050 tables and still growing every week.
Enterprise software has gone wrong in two opposite directions. Fortresses (SAP, Oracle, Workday) dictate workflows; modern SaaS archipelagos (Notion, Linear, Slack) are delightful alone but isolated together. Most companies pick a side. We started building the bridge.
Internal beta covers AlphaCore (contracts), Automation, FyBrain (intelligence), and FyDrive (files). Materials and Finance landed in July 2026 with 125 and 92 permission-gated API routes respectively. Public access is gated behind the waitlist. By March 2028, Fyboard will replace SAP at the parent group — that is the public bet, and if we miss it we will publish the post-mortem.
For more on the founding argument, read /origin. For the principles, read /manifesto. To apply to build with us, see /careers.
Fyboard is the operating system built to run an Indian EPC company with ₹1,300 crore+ in FY25-26 revenue.
Engines, not apps. Forged in concrete. Refined by code.
Every contract, file, form, document, person, approval, and automation run inside Fyboard is the same kind of thing — an operational record. Modules don't integrate. They compose, because they all speak the same vocabulary.
"Once everything is a record, modules don't need bespoke integrations to talk to each other."
Fyboard architecture · principle 01
Each module does one job, well, over the same operational records. 4 are on internal beta; 7 are still being built. None are public yet.
Contract lifecycle — obligations, eSign and renewals.
Drafting → negotiation → eSign → renewal — every clause is a record.
Cloud storage — files, versions and lineage.
Drawings, scans, signed PDFs — lineage-safe, permissioned, versioned.
AI processing — OCR, embeddings and semantic search.
OCR, extraction, summarisation — in-region by default, external LLM calls off.
Workflows, triggers and business processes.
Approvals, SLAs, escalations — one engine across every module.
Relational data system — schema and forms for intake.
Vendor intake, employee onboarding, audit checklists — typed, queryable, governed.
Productivity documents — collaborative, block-level.
Memos, SOPs, RFCs — versioned, citation-aware, AI-assisted.
HRMS — directory, attendance, payroll, compliance, and the org spine the rest reads from.
Roles, permissions, reporting lines — the spine other modules read from.
Item master, units of measure, business partners, vendor catalog and inventory.
Item master, UoM, business partners, vendor catalog, stock movements.
◆ 125 permission-gated API routes
General ledger, payables and receivables, payments, treasury and financial reports.
GL accounts, payables and receivables, payments, treasury, an append-only journal.
◆ 92 permission-gated API routes
Projects, portfolios, delivery, resources, budgets and cost control.
Portfolios, delivery plans, resources, budgets, cost control against real commitments.
Integration hub for connected apps, accounts and scoped external data.
Connected apps, accounts and streams — external data scoped by the same permissions.
12 engines carry 11 modules. Permissions and Audit sit underneath all of it — not as an admin page bolted on late, but as the layer every other engine has to ask before it can read or write a record. Improvements propagate to every surface, instantly.
Four-tier organisation RBAC with multi-axis resolution on every request, plus dimensional row scoping and time-bounded grants.
Mutation auditing across contracts, materials and finance — actor, action and before/after diff, written append-only.
Intelligence — OCR, extraction, embeddings and semantic search. Also ships its own module surface.
Orchestration — approvals, routing, conditions, escalations and SLAs.
Storage — files, versions, lineage and per-record file permissions.
Composable reporting over the same records every module writes.
Mapped, validated, resumable ingestion into any record type.
Record-table engine — schema-driven columns, filters, editing and bulk actions.
Dynamic fields and forms rendered from a schema definition.
Calendar over records.
Spreadsheet where cells are records, not values.
Computed values across records.
12 engines power 11 modules. The same comment is the same comment everywhere. The same approval is the same approval. The same permission check answers for all of them. There is no duct-tape between products because there is nothing to tape together.
Fyboard is built inside the digital arm of an Indian EPC group with ₹1,300 crore+ in FY25-26 revenue. One person writes it, schema-first, with no marketing-led roadmap. Slow on features, fast on foundations.
I'm inside the company I'm building for, not adjacent to it. That changes what ships and what I refuse to ship. Every engine is documented and every permission is generated from a checked-in catalog, so the platform does not live only in my head.
Most AI-first platforms quietly route your records through a global LLM API. We don't. Fyboard runs the intelligence layer inside the same region as your storage layer — and inside your tenant.
Our AI engine sits inside your cluster, in your region. No prompt or document is shipped to an external LLM provider.
Default tenancy is in India. Other regions are configurable. Where your data lives is a deployment decision, not a guess.
Files in FyDrive, records in DataBoard, contracts in AlphaCore — all stay inside the boundary your IT team draws.
Built this way because the same regulators we work with on highways and power grids audit our software. Compliance isn't bolted on — it's the floor we started from.
Solo operators don't have to graduate to a different tool when their team grows. The records, the workflows, the files — they all stay. Only the surface area changes.
FyDrive: secure links, granular permissions, lineage you can audit.
AlphaCore: send, sign, store. Every clause is a record.
FyBrain: drop in PDFs and FyBrain extracts the structure.
Automation: form → folder → notify → archive, without glue code.
The freelancer's signed PDF and the enterprise's procurement chain are the same kind of record.
Fyboard is in internal beta. 4 modules ship for internal teams and close testers right now. 7 are in active development, including Materials and Finance, which landed in July 2026. Public access is gated behind the waitlist on purpose — we open modules when they're stable, not when the calendar says we should.
The engines hold real production data — contracts, audits, signatures. Schema stability and migration safety come first. We'd rather wait six months than ship something that locks our users into a shape we'd later regret.
PDF tools, OCR, file converters, a lightweight eSign. Free with fair usage caps, opening before the modules. They're the easiest way to feel the engines without committing to the platform.
See current toolsFour horizons. We move forward when the previous one is stable, not when the calendar says so.
AlphaCore, FyDrive, FyBrain and Automation run on internal beta. Materials and Finance landed in July 2026 — 125 and 92 permission-gated API routes — built on primitives that did not have to change to accommodate them. DataBoard, FyDoc, People, Project System and FyIntegrations are hardening alongside.
Free utilities open first. Modules open one by one once schema stability is proven outside the parent group.
Target horizon for replacing the current ERP stack end to end. If we miss it, we will say so plainly, with the post-mortem published.
Grouped into four conversations: what the platform is, how the trust model works, where it stands, and what we're committing to.
A records platform with eleven modules on top. Every contract, file, form, document, person, approval and automation run is the same kind of thing — an operational record — which is how the modules compose without bespoke integrations.
Read the definitionA row is a database concept — fields in a table. A record is an operational concept — it carries ownership, history, context, automation hooks, intelligence, and permissions. Modules read and write the same records, so a contract in AlphaCore is the same kind of object as a memo in FyDoc or a file in FyDrive.
See the records spineSaaS suites are integrated with duct tape — each app has its own model and the seams show. ERPs force your operations into their model. Fyboard is records-first: every module reads and writes the same operational records, so composition is native, not bolted on.
In-region by default: your storage and compute sit in the same region as your tenant, India by default, other regions configurable. External LLM calls are off by default — no prompt or document leaves for a third-party model provider. In-tenant FyBrain deployment, where inference runs inside your own cluster, is available for enterprise agreements.
Region & sovereigntyPermissions are four-tier — module gate, role permissions, derived module/feature permissions, and per-user overrides — resolved on every request across multiple axes: capability, dimensional row scope, and per-record access. Row scoping is hierarchical (cost centre, business unit, region, with descendants) and grants are time-bounded with effective-from and effective-until windows. Eighteen Postgres functions back the resolution, and in Materials an out-of-scope direct fetch returns 404 rather than 403, so record existence itself is not leaked. Mutations across contracts, materials and finance are audited with the actor and a before/after diff. We hold no third-party security attestation yet — SOC 2 and ISO 27001 controls are in build, with no committed dates.
Where we actually areYes. Records are typed and queryable, files come with their lineage, and contracts ship with their full audit trail. Export is a first-class operation, not a customer-success ticket.
Because the engines hold real production data — contracts, audits, signatures. Schema stability and migration safety come before broader release. We open modules when they're stable, not when the calendar says we should.
Live for internal teams and close testers: AlphaCore, FyDrive, FyBrain, Automation. In active development: DataBoard, FyDoc, People, and the two newest — Materials and Finance, which landed in July 2026 with 125 and 92 permission-gated API routes respectively. Public access for all of these is gated behind the waitlist.
See full statusPDF tools (merge, split, compress, convert), image + OCR utilities, file converters, web scraping tools, and lightweight eSign. Free with fair usage caps. They open before the modules — the easiest way to feel the engines without committing to the platform.
Try the toolsFyboard is built inside the digital arm of an Indian EPC group with ₹1,300 crore+ in FY25-26 revenue. It is written by one person, deliberately, schema-first, inside a real operating company.
Read the originBy March 2028, the parent group replaces SAP with Fyboard. That's the public bet. If the platform can't run real infrastructure work end-to-end, it isn't ready for anyone. If we miss it, we will say so plainly, with the post-mortem published.
The long viewWe answer everything that lands in our inbox — usually within a working day. Build context with us; it sharpens what we ship.
Fyboard is in invite-only beta. We're growing deliberately, proving quality with every user before scaling. If that's the kind of thing you'd want to build on, we'd like to hear from you.