Onixa is a lab of one. I build working software for businesses: revenue tooling, AI pipelines, workflow automation, internal systems. You bring the problem. The output is code that runs, not a deck.
Sales intelligence. Territory logic. Account scoring and enrichment. Sequence automation. Call prep that actually reads the account.
I build this for the operator running the motion, not the manager auditing it. I’ve sat in the seat. I know what the pipeline view leaves out.
Multi-step pipelines that ingest, reason, and act. Built on MCP, Claude, or whatever model fits the job.
Not ChatGPT wrappers. Pipelines that do real autonomous work and hand back structured output with a review gate where a human should still look.
Work that currently runs on human hours, rebuilt as pipelines. Research, enrichment, qualification, reporting, briefing.
You stay in the loop where judgment matters. The grunt work disappears into infrastructure.
The software your team actually needs, but isn’t on any vendor’s roadmap and isn’t worth a hire.
Scoped tight. Built fast. Delivered running. It sits on the stack you already have and the documentation explains itself.
I start by finding the actual problem. Not the symptom. Not the feature list you arrived with. The friction underneath. From there: a scope, a timeline, a defined output. No discovery phase that drags on for six weeks before anyone writes a line of code.
The output is working software, not a strategy document. Every engagement ends with code that runs, documentation that explains it, and a walkthrough that makes it yours. You operate it. I don’t stay retained to run it for you.
One operator with AI erases the coordination tax that eats 40% of a team’s calendar. The context stays in one head. Decisions are instant. What takes four people six weeks takes me two to three, and the quality is higher because nothing gets lost in the handoff.
If the machine keeps running after I’m gone, that’s the point.
That’s what it means to hand it back.
I built it because the tool I wanted didn’t exist. A canvas where accounts are objects, time is an axis, and the whole workspace is readable by any connected model. No translation layer. No summarisation step eating the signal.
MCP server with 12 tools. Account scoring and enrichment. Sequence composition with outcome attribution. Call prep drawn from public signal. Scoring diff and topology views. I use it daily.
▸ silvercontext.comRevenue intelligence pipeline for a mid-market sales team. Custom enrichment DAG, scoring model, Slack briefing layer. The reps stopped opening their CRM first.
Workflow automation for a professional services firm. Document ingestion, classification, and routing. Cut manual processing time by roughly 85% and freed a full seat.
One paragraph over email. What’s broken, what you’ve already tried, what outcome actually matters. No NDA at this stage. I’m not going to run off with your business idea.
45 minutes on a call. I ask the questions that matter. The goal is to find the root problem, not stenograph the feature you showed up asking for.
What gets built. What it does. What it deliberately doesn’t. Timeline, delivery format, and a clear definition of done. Signed before a line of code.
I build. You stay available for questions I can’t answer without you. One check-in at the midpoint. No weekly standups, no status decks.
Working software. Documentation. A walkthrough that gets your team operating it without me. I stay reachable for questions for 30 days after delivery.
Typical engagement runs 2 to 6 weeks depending on scope.
I take one engagement at a time. If the slot is full, I’ll tell you.
Send a paragraph. Not the solution. The problem. What’s costing time, money, or accuracy. What you’ve already tried and why it didn’t hold.
If it sounds like something I can build, I’ll reply. If it doesn’t, I’ll say so and point you at someone who can.