- Software vendors
- Fixed price in 3 days
- First feature in 1-3 weeks
AI features for your software,
built inside your repo.
An in-app assistant, an MCP server for your product, or an integration into the systems your customers run. A senior engineer works in your codebase, your reviews and your roadmap.
You talk to the engineer who builds it.
12products shipped
7 built solo, 5 embedded in client teams.
What gets built.Features and integrations for your product.
- 01
In-app assistant
Users ask in plain words and the assistant acts through your own API, with their permissions and a confirm step before every write. The pattern in detail: the in-app assistant.
- 02
MCP server for your product
Your product as tools for Claude, ChatGPT or Cursor, read-first and designed for agents rather than one tool per endpoint. How the MCP server is built.
- 03
Document intake
Orders, invoices, referrals and forms read into your data model, checked against your rules, with a review queue for anything uncertain. Examples: freight order entry and clinic referral intake.
- 04
Integrations to your customers' systems
Connectors to the ERP, TMS, practice system, accounting tool or CRM your customers already run, built against their APIs and limited to the scopes they grant. See the SaaS builds.
Every feature ships withpermissions per useran eval harnesslogs and cost trackinga fallback path
Who it is for.And who should not hire me.
- 01
Software vendors with a full roadmap
Vertical SaaS in logistics, clinic and practice, insurance and accounting, plus B2B SaaS and software houses. Your engineers are booked, and the feature needs model behaviour nobody on the team has shipped yet.
- 02
Agencies under their own brand
Agencies that sell an AI feature to a client and need a senior engineer to build it. The work carries your name, comes with a written handover, and can be maintained by a third engineer.
- 03
Teams that already have an AI team
If your engineers have shipped LLM features, build it yourselves and keep the knowledge in-house. I will say so on the call. What helps then is one bounded feature, or a design review before it reaches the backlog.
Inside your team.Your repo, your review, your roadmap.
- 01
Your repository
Commits land on your branches and pass your CI. The code is yours from the first commit, in your repo, with no part kept back.
- 02
Your review
Pull requests go to your reviewers and merge through your release process. Nothing ships that your team has not read.
- 03
Your roadmap
You rank the tickets. Scope and trade-offs are agreed with your product lead, with a written scope before the build and a weekly demo after it.
- 04
Your data and an NDA
An NDA before the detailed discussion if you want one. The work runs in your environments, and I do not keep copies of customer data.
- 05
Your name, if you resell
For agencies and resellers: the work goes out under your brand, with a handover your own team can run.
Shipped, and public.Check the MCP server yourself.
- How to add the MCP server
The public MCP server at ergini.com/mcp
Six read-only tools: list_services, find_use_cases, get_use_case, estimate_project, list_case_studies and book_call. Stateless, no sign-in, and only public site content. Add it to Claude, Cursor or any MCP client.
Products in production
Caldra reads the inbox and proposes meeting times for approval. AURA Studio turns one garment photo into on-model images and video. Xandidate reads every applicant against the job post, with the reasons behind each score.
Call to launch.In weeks, not quarters.
Start with the callErgini · Software & AI developer
I build every project myself,so I only take a few at a time.

3 days
A fixed price, in writing,
three days after the call.
- Scoped from what we discuss on the call.
- Agreed before any work starts.
Builds like these.Worked examples, end to end.
All use casesA remote MCP server so customers can use your SaaS from Claude, ChatGPT and Copilot
Your product inside Claude, ChatGPT and Copilot Studio: OAuth sign-in, a small set of agent-shaped tools, per-tenant audit logs and rate limits that protect your API.
Your REST API / OAuth provider (Auth0, WorkOS or your own) / MCP TypeScript SDK
An AI bill growing faster than your users: a cost audit, model routing and budgets
Traces model spend to each feature and customer, fixes caching, batching and prompt bloat, routes simple calls to small models, and proves quality holds with evals.
OpenAI and Anthropic APIs / Langfuse or Helicone / AI gateway
An in-app AI assistant that acts on the user's own data, not one that only quotes the docs
Tool calling over your own API with the signed-in user's permissions, a confirm step before every write, undo, per-plan usage caps and a trace of every step.
Your product's API / Vercel AI SDK / OpenAI or Anthropic
Questions.Answered straight.
Do you work inside our repository and our team process?
Yes. Commits land on your branches, pass your CI and go through your reviewers and release process. Planning follows your roadmap and your tools, not a separate project board I bring with me.
What is the difference between an in-app assistant and an MCP server?
The in-app assistant lives inside your product and acts for the signed-in user through your API, with a confirm step before any write. An MCP server exposes your product to outside AI clients such as Claude or Cursor, with tools designed for agents and the same user permissions. Many vendors need both, and they share one tool layer.
Can you build an MCP server for our product?
Yes. The tools are designed around what an agent needs to answer and act, not one tool per endpoint. Reads come first; writes stay behind an approval step. The public example is ergini.com/mcp: six read-only tools, no sign-in, and only public site content.
Do you white-label the work for agencies?
Yes. The work can go out under your brand, with a written handover, documentation a third engineer can maintain, and an agreed rule for who speaks to the end client. We settle those terms before the first sprint.
Who owns the code?
You do. The code lives in your repository from the first commit, and nothing is kept back as a private component of mine.
Will you sign an NDA before we share the product details?
Yes. An NDA can be signed before the detailed discussion. The build runs in your environments, and I do not need or keep copies of customer data.
When should a software company build this in-house instead?
When engineers who have already shipped model features are free and the work is a roadmap item rather than a gap. In that case a design review before the backlog is often worth more than a build, and I say so on the call.
Pick a time.The call is free.
30 minutes, on Google Meet
Prefer email? ergin@ergini.com. I reply within hours.
See it built: the Caldra case study · the Xandidate case study
Related: AI integration · AI agent development · AI for B2B SaaS · AI for healthcare