ALTHERIX Talk to an Engineer
Talk to an Engineer
contact@altherix.in+91 81529 23515LinkedIn
Products

What we're building.

Products that come out of the problems we solve, for clients and for ourselves, built with the same engineering discipline.

In use · Early accessCodeGate by AltherixRank candidates on the code they write, using hidden tests and AI grading.In developmentMandate by AltherixA safe way to connect AI agents to the systems your business already runs on.
In use · Early access

CodeGate by Altherix

Rank candidates on the code they write, not the CV they send.

CodeGate is an online coding assessment. Candidates solve real-world problems in a timed browser editor, and hidden tests plus AI grading turn every answer into a score you can rank.

Try CodeGate

Built for our own hiring rounds. JavaScript today, more languages planned.

Private, timed tests

Each candidate gets a single-use link by email. The timer runs on the server, work autosaves, and answers submit at time-up.

Tests decide correctness

Hidden test cases score whether the code works. The AI can't change that part of the score.

AI grades the rest

Edge cases, code quality and efficiency are scored against a written rubric, with a short summary of each answer.

Signals, not guesswork

Tab switches, pastes and copies are logged, and follow-up questions are generated about each candidate's own code.

CodeGate candidate screen: the question and rules on the left, a code editor with the candidate's solution on the right, a countdown timer and a Submit button.
Candidate view: timed editor with sample tests (demo data)
CodeGate admin screen: test status, candidate counts, average score out of 20, an invite form and a ranked candidate list with scores, flags and decisions.
Admin view: invite, track and rank candidates (demo data)
How CodeGate scores

Ten points per question. The tests decide what works.

4Correctness, from hidden tests
2Edge cases
2Code quality
2Efficiency

Correctness comes only from hidden tests. AI scores the other six points against a written rubric.

  1. 01InviteAdd candidates. Each one gets a private test link by email.
  2. 02TestCandidates solve real-world problems in a browser code editor, with sample tests they can run.
  3. 03GradeOne click runs the hidden tests and AI-grades every answer.
  4. 04ShortlistRank candidates by score, review the flags, and use the generated questions on a short follow-up call.
Illustration: business systems on one side and an AI agent on the other, connected through a gate where some requests pass and one is held for approval.
In development · Early access soon

Mandate by Altherix

Your ERP is about to get an AI colleague. Who decides what it can do?

Mandate is a safe way to connect AI agents to the systems your business already runs on.

Request early access

Or email contact@altherix.in with the subject “Mandate”.

Works with what you run

ERP, SQL databases and in-house applications, made usable by AI agents.

Your permissions, not new ones

Agents see only what their user is allowed to see.

A person approves risky actions

Writing data, sending documents or moving money waits for approval.

Every action on record

Each request and change is logged and can be replayed.

How Mandate works

AI that works inside your systems, only within its mandate.

  1. 01An AI agent asksAn assistant wants to read a record or take an action in one of your systems.
  2. 02Mandate checksIs this user allowed? Does this action need approval? The request is logged either way.
  3. 03Your system respondsOnly permitted, approved actions reach the system. Everything else stops at Mandate.

Want to use one of these early?

Tell us which product you're interested in and what you'd use it for. An engineer will reply.

Try CodeGateMandate early access