Built in India, for how India writes
Every one of those went straight past a spell checker. Sahi catches them — and leaves Telmisartan, NABH and Dr Kulkarni alone.
Self-hosted. Your server, your data, your dictionary.
The gap nobody else fills
These are not typos. They are the habits that are invisible from inside India and immediately obvious to anyone reading your correspondence from outside it.
It knows your words
Telmisartan, Pantop, Combiflam, Shelcal, Augmentin — and anything else you dispense, added in a minute from the admin.
OPD, IPD, NICU, HbA1c, USG, TPA, ABHA, NABH read as terms, not as spelling mistakes.
Doctor names, ward names, your hospital's own terminology. A word one person adds is remembered for the whole organisation.
Write the way you think, in Roman Hindi or a mix, and get back professional English that says exactly what you meant.
Everywhere your staff write
One dictionary, one set of rules, one place to manage all of it.
Discharge summary, diagnosis, notes, prescription, remarks. One script tag on the screen — no change to how anyone works.
A browser extension that underlines as you pause. Deployed across every machine through standard Chrome policy — nothing for staff to set up.
Letters and summaries written in Word, and mail composed in the desktop client, not just the web one.
Select text in any app and correct it. Sahi sees only what you highlighted — never anything else you type.
Honestly compared
Grammarly is an excellent product with a decade of engineering behind it. It is also built for a different reader, in a different country, on somebody else's servers.
| Sahi | Grammarly | |
|---|---|---|
| Indian English habits | Built in — 60+ rules for revert back, do the needful, informations, prepone | Not covered |
| Clinical and pharma vocabulary | Ships knowing it, and your own terms are added in a minute | Underlines most drug names until each user adds them individually |
| Hinglish to English | Yes — one click | No |
| Where your text goes | Your server. On your own hardware if you want, with AI switched off entirely | Their cloud, outside India |
| Works inside your HIS | Yes — one script tag, including Telerik editors | Browser only |
| What administrators see | Repeat mistakes and acceptance rates, by department and by person | Seat usage and generic scores |
| Billing | In rupees, on your AMC cycle, invoiced in India | In dollars, card only |
| Long-form style and tone depth | Good — rewriting for formal, friendly, concise | Stronger. A decade of work on English prose style. |
| Consumer free tier | No — this is sold to organisations | Yes |
Data residency
Sahi is self-hosted. With the AI option off it runs entirely on the rule pack and a local grammar service on your own infrastructure — and you can write that into the contract.
Message text is never kept beyond a short cache you can switch off completely. Reports count phrases, never messages.
Password fields are excluded outright. Any field can be excluded. Sahi can be turned off per site, per department or per person.
One access key per person. The day somebody leaves, revoke it and every copy of Sahi they hold stops working immediately.
Before anything is counted, phrases over six words or containing a number are dropped — so MRNs, amounts and doses never reach a report.
For the people running it
A 400-bed hospital is not one team. Sahi has four levels, so a ward manager sees their ward, an administrator sees the hospital, and nobody sees more than they should.
Creates hospitals, sets plans and monthly ceilings, switches AI on or off per tenant, and sees usage across all of them.
Departments, staff, keys, the dictionary and house-style rules — plus the full picture of what the hospital keeps getting wrong.
Sees only their own department: who is writing, what they repeat, and whether corrections are being taken. A coaching list, not a scoreboard.
Signs in, collects their own key, sets up their own devices, and sees their own progress. No IT ticket required.
Pricing
Invoiced in India, on your AMC cycle if you already work with us. We are running a small free pilot while the product settles — ask, and we will tell you honestly whether you fit it.
Questions we actually get asked
Only if you choose to switch the AI option on, and even then only the text in the field being checked, which is processed and returned immediately rather than stored or used for training. Switch it off and Sahi runs entirely on your own infrastructure — the rule pack and a local grammar service — and nothing leaves at all. Most hospitals start there.
Yes. Sahi is self-hosted by design, and there is a container stack that brings up the application, the database and the grammar service in one command. It runs happily on a modest virtual machine, with or without internet access.
Yes, including screens built on ASP.NET with server-generated control IDs and Telerik editors. It is one script tag on the page, and it survives partial postbacks. Nobody has to change how they work.
The field selectors are published from the admin, not baked into the extension. When a site changes its markup, a corrected selector reaches every installed copy within half an hour — no app store review, no reinstall.
It is the reason several of them will like Sahi. They write the way they think, in Roman Hindi or a mix, and get back professional English that says exactly what they meant — with nothing added that they did not write.
No. Reports count phrases, never messages, and anything longer than six words or containing a number is discarded before it is stored. A manager can see that their team writes "revert back" often; they cannot see anybody's correspondence.
A pilot with a handful of people runs the same day. A hospital-wide rollout is mostly deciding who gets it, adding your own vocabulary, and pushing the extension through Chrome policy — usually a week, most of it waiting on your IT change window.
Pick the team whose writing goes outside the hospital most often. Two weeks is enough to see whether it is helping — and the report at the end of it will tell you plainly.