Sales
Win the mandate before you work it.
- Leads board
- Openings & contacts
- Bulk lead import
- Shared email inbox
- Sequenced follow-ups
AxureOne CRM · early access
The AI-native CRM for recruitment agencies.
Not an applicant tracker with a chatbot bolted to the side. A recruitment platform where an agent takes a brief, screens hundreds of profiles for free, spends your sourcing budget only where it earned the right to — and hands your desk a ranked call list you can audit line by line.
Built in India, for how Indian recruitment actually works. Self-hostable. Running a real agency desk every day — ours.
sourcing run · live transcriptrunning
312
screened
4
spent on
11
call list
- 3portals, one database
- 21permissioned modules
- 10white-label résumé templates
- 0external keys needed to run it
The premise
Most recruitment CRMs are a filing cabinet with a search box.
They are very good at recording what your recruiters already did. That was a fair deal when nothing else was possible. It isn’t any more — and a “summarise” button on a candidate profile is not the answer to it.
Records what your recruiters already did.
Does a share of the work itself — then shows you exactly how it did it.
“AI” means a summarise button on a profile.
An agent with tools, a budget, memory of what it already tried, and a transcript you can read.
Built for the US or UK. India localised in later.
CTC in lakhs, notice buyouts, GST, DPDP and DLT — in the schema from the first migration.
Your candidate database lives in someone else's cloud.
Self-hostable on your own Postgres. Every integration stays dormant until you key it.
The agent
How one sourcing run actually goes.
An assistant answers when you speak to it. This takes a role, works it for minutes on end while nobody watches, decides where money is worth spending, and writes down every decision it made on the way. Six steps, in the order they happen.
- 01
It takes a brief
Before it spends anything, the agent drafts its own intake questions for the role and asks the recruiter what a real fit looks like — and how many calls they actually want. The answers are frozen into the run, so an edit to the JD halfway through can't move the goalposts.
› intake · 6 questions drafted · target: 12 calls · brief frozen
- 02
It writes its own searches — then rewrites them
One keyword and one page of results is a guess, not a search. The agent runs query variants, reads the total counts, and changes its own approach when page one comes back as noise. Generic title, skill-led, location-relaxed: it decides, and it says why.
› search · variant “skill-led” · 1,240 matched · page 2 requested
- 03
It screens hundreds before spending a rupee
A locked profile already carries headline, company history, years of experience, current and expected pay, skills, education, relocation cities, notice availability and last-active date. That is enough to judge fit properly. Hundreds get screened for free, in batches, and banded strong / stretch / thin / no.
› screen · batch 25 · 4 strong · 7 stretch · 14 no · ₹0 spent
- 04
It spends only where screening earned it
Revealing a contact costs real money, so the agent buys them only for profiles that cleared the bar — under a hard per-run ceiling that only a super admin can raise, and a pay ceiling that keeps it from chasing people the role could never afford.
› fetch · 4 contacts revealed · 21 of 25 run ceiling remaining
- 05
It hands over a call list, not a spreadsheet
The run ends with candidates ranked, each with talking points drawn from their actual profile, imported straight onto the job's pipeline. If the market genuinely doesn't hold enough people, it says so and tells you what it relaxed to get close.
› finish · 11 on call list · imported to pipeline · market note attached
- 06
It shows its work — or it stops
Every turn is written down: the thought, the tool it chose, what came back, what it cost. Runs are closable, resumable, and budgeted. And when the model can't be reached, the run blocks honestly instead of quietly degrading into “no candidates fit”.
› transcript · 31 steps · 18 model calls · run resumable
AI where it earns its place
Twelve jobs it does, not twelve slides about it.
None of these are a chat window. They are ordinary parts of the desk that happen to be done by a model, sitting behind the same buttons a recruiter was already pressing.
- 01
Résumé parsing
Drop a CV in any shape and get structured fields back — name, contact, headline, experience in months, CTC as absolute rupees, notice, skills, roles, education.
- 02
Résumé OCR
Some résumés arrive as rendered images with no text layer at all. Those get read page by page with vision into exactly the same structured shape.
- 03
Candidate summaries
A factual two-to-three sentence summary for a client submission, written only from what is actually in the profile. No embellishment, no invented achievements.
- 04
Client pitch
Why this candidate, for this role, at this company — their real facts mapped to the role's real needs, with the strongest one named.
- 05
Fit ranking
A whole shortlist scored against one role in a single batched pass: 0–100 plus a one-line reason each, layered on top of a deterministic score that works with no AI at all.
- 06
Outreach drafting
A short, warm first-touch WhatsApp message or email that references one real thing from the candidate's profile — not a mail-merge with their name in it.
- 07
Job descriptions
A rough brief or a pasted JD becomes the whole new-job form: title, experience band, work mode, location, CTC range, and the description sections.
- 08
Interview scorecards
Raw interview notes become a structured scorecard against the role, so feedback survives the day it was given.
- 09
Lead contact discovery
Finds who to call at a target company by reading the open web — and every contact must carry the page it was read from, or the parser drops the row.
- 10
Lead outreach
The opener for a lead, grounded in the actual opening it was captured from: a call script, an email, or a WhatsApp message.
- 11
Careers-page extraction
Reads a company's job page and fills only the gaps structured data missed — it is never invited to overwrite a fact the page stated outright.
- 12
Search query building
Turns a role into the keyword query a talent database will actually return people for, while the numeric filters come deterministically from the job's own fields.
Every one of them is optional. Turn the AI layer off and the platform keeps working — fit scoring falls back to a deterministic, unit-tested score, outreach falls back to a written template, and nothing anywhere throws.
The whole desk
Everything between the lead and the invoice.
A recruitment agency does not stop at “candidate submitted”. It chases the mandate, signs the agreement, tracks the guarantee clock, raises the GST invoice and pays the recruiter their cut. All of that is in here, in one database.
swipe →
Three sides
Your team, your clients and your candidates — in one system.
Not three products that sync overnight. One database, one login system, three views of it — so a status a recruiter changes is the status the client and the candidate see.
Your team
Recruiter console
- The whole desk — sales through invoice
- Command palette on ⌘K
- Per-stage ageing, so nothing rots quietly
- Every list bounded; a 25k pool stays fast
The companies you bill
Client portal
- Review submitted candidates on a branded résumé
- Rate, comment, and decide in one place
- Request a new role without emailing anyone
- Or skip the account entirely — a single-use review link needs no login at all
The people you place
Candidate portal
- Browse and apply to your live roles
- Track every application's real stage
- Keep their own profile and documents current
- Consent and erasure they control themselves
the part that wins the deal
A hiring manager who must create an account before they can say yes is a hiring manager who says yes on Thursday instead of Monday. So they don’t have to: every submission can go out as a single-use review link that opens the branded résumé, takes their decision and their feedback, and needs no login at all.
Built here, not localised later
The India bits are in the schema, not in a custom field.
Every global ATS eventually offers you a text box called “Notice period”. None of them will calculate a buyout, split fixed from variable, or raise an invoice your CA will accept.
CTC the way India quotes it
Lakhs per annum, with the fixed and variable split kept apart, stored as exact decimals — never a float that quietly loses ₹1 on a ₹18L package.
Notice periods and buyouts
Notice in days and the buyout amount are first-class fields, because they are the two numbers that decide whether an offer closes.
GST invoicing that files
18% GST on SAC 998512, GSTIN on the invoice, a proper invoice counter, and IRN e-invoicing through a GSP when you switch it on.
DPDP, not a cookie banner
Consent recorded per purpose, a real data export, a real erasure path, and retention cut-offs that actually run.
TRAI DLT and WhatsApp opt-in
Entity and sender IDs on SMS, and no business-initiated WhatsApp message before the candidate has opted in.
Aadhaar eSign on agreements
Client agreements and offer letters signed in-platform, in-house or through an ESP, with an audit certificate attached.
Asia/Kolkata everywhere
The whole platform runs on IST. UTC made “due today” off by one, once. It does not any more.
Payslip-read CTC verification
Verify a claimed package against the payslip instead of taking the number on trust — one of the two things that break an Indian offer.
Control
Your database. Your keys. Your infrastructure, if you want it.
A recruitment CRM holds the most sensitive thing your business owns: everyone’s salary, everyone’s phone number, and the fact that they are quietly looking. This one is built like that matters.
Authentication we own
Argon2 password hashing, opaque session tokens where only the SHA-256 hash is ever stored, TOTP two-factor with backup codes. No third-party identity provider sits between you and your own database.
Permissions enforced twice
Twenty modules, each none / view / edit, resolved as per-user override, then group, then role default — checked in the page layout and again inside every single write. Hidden UI is never the boundary.
Redaction a template cannot leak
There is exactly one way to build a résumé, and it removes hidden fields on the way in. A template does not receive the candidate's phone number and then choose not to print it — it never receives it.
Impersonation you can audit
Staff can view the app as a client or candidate to debug what they are seeing. The authority for that lives in the session row on the server, never in a cookie, and every use is logged.
Dormant until you key it
Email, WhatsApp, SMS, telephony, storage, e-sign, e-invoice and AI are all switched off until you enter credentials in-app, encrypted at rest with AES-256-GCM. A fresh install builds, migrates and runs with nothing at all.
Yours to host
Postgres and Docker. No mandatory external SaaS anywhere in the core. If you want it on your own infrastructure, that is the default, not an enterprise add-on.
The rules it runs by
The hard part of AI in recruitment isn’t the clever answer. It’s the wrong one.
A model that invents a phone number, or quietly returns a thin shortlist because it couldn’t reach the API, costs you a client — and you find out weeks later. So the platform is built around five rules it is not allowed to break.
rule 01
A failed AI call never becomes a confident answer.
Every AI function returns nothing on failure and never throws. The feature falls back to what it did before AI existed — a deterministic score, a written template — rather than shipping a guess into your client's inbox.
rule 02
A blocked run says it is blocked.
If a sourcing run cannot reach the model, it retries once and then stops and tells you. It will not hand you a thin shortlist that looks like a market verdict.
rule 03
Every discovered contact carries its source.
A model willing to invent a phone number will happily invent a citation too. So the citation is required by the parser, not by the prompt: a contact with no page behind it is dropped before anyone sees it.
rule 04
“Queued” never gets rounded up to “sent”.
If email isn't configured, the platform says the message is queued. A recruiter must never believe a client was emailed when they were not — that is how a placement dies quietly.
rule 05
The pipeline and the client's view stay in step.
What a client sees and what your board counts are reconciled by one piece of code, called by every writer. There is no path where a candidate is submitted in one screen and missing from another.
We’re customer number one
We didn’t build this to sell it. We built it because we needed it.
AxureOne is a boutique recruitment firm in Ahmedabad, working roles across India. Every part of this platform exists because a real search was slower than it should have been, or a real invoice was raised in a spreadsheet, or a real client waited two days for a login they didn’t need.
That is also why it has a 90-day guarantee clock and a commission ledger in it. Nobody adds those to a demo.
Hiring rather than buying software? See how we recruit — transparent fees from 6.5% or ₹6,499, invoiced only when someone actually joins.
- 90 daysfree replacement guarantee*, tracked in the platform
- 6.5%of CTC, or ₹6,499 flat — published, not negotiated per client
- ₹0invoiced until the candidate joins
- Pan-Indiaroles worked from an Ahmedabad desk
Questions
The things agency owners actually ask us.
Anything not answered here, ask us directly — you’ll get a straight answer, including “not yet”.
It is in early access. The platform is live and running a real recruitment desk every day — ours — and we are opening it to a small number of agencies we can support properly rather than to everyone at once. Get in touch and we'll show you the actual product, not a slide.
We haven't published pricing yet, and we would rather say that than invent a number on a webpage. Early-access conversations set it, and whatever we land on will be published the same way our recruitment fees are.
Yes. It runs on PostgreSQL and Docker, and a fresh install builds, migrates, seeds and runs with no external credentials at all. Every integration — email, WhatsApp, SMS, calling, storage, e-sign, e-invoice, AI — is dormant until you key it in the app.
Anthropic's Claude, driven through a subscription or key you provision inside the platform yourself. The AI layer is a single module with one contract, so the model behind it is a configuration decision rather than an architectural one.
No. Your database is yours, it lives in your Postgres, and nothing in the platform ships candidate data anywhere to improve a model.
Yes. There is bulk candidate import — including from Zoho Recruit — plus CV import that parses each file into structured fields, and duplicate detection so a migration doesn't leave you with three of everyone.
An assistant answers when spoken to. This runs a task end to end on a budget: it plans, searches, screens hundreds of profiles, decides where money is worth spending, and writes down every step so you can audit the judgement rather than trust it.
It was built by a boutique agency for a boutique agency's economics, which is why it has commissions, GST invoicing and a 90-day guarantee clock in it rather than an enterprise implementation project. Modules you don't grant to anyone simply aren't there.
Early access
See it running on a real desk.
We’re opening AxureOne CRM to a small number of recruitment agencies we can actually support — not to a waitlist we’ll never work through. Tell us how your desk runs today and we’ll show you the live product, including the parts that aren’t finished.
- A walk through a real sourcing run, transcript and all
- A straight answer on what it does and doesn't do yet
- Whether your data would live with us or with you
Or email us directly at hello@axureone.com.