Software Engineering · Agent Systems · Consulting
Agent-built software for places the cloud can't go
I'm a software engineer who builds the tools people ask for. I build them by managing a fleet of coding agents across many models, and I ship them to run offline, on your hardware, with your models.
Services
What I do
Agent Systems & Multi-Model Pipelines
Multi-agent workflows that produce finished work, not demos: writer, reviewer and auditor roles, bounded loops, and a human review gate where it matters. Bring your own models—open weights on your hardware, or frontier APIs—and mix them per task.
Offline & Airgapped Deployment
The same software, running where there is no internet. Self-contained bundles, offline package management, signed single-file builds, and models that live on your hardware. Nothing leaves the device unless you decide it does.
Custom Tooling & Full-Stack Development
The tool you actually asked for: web apps, APIs, CLIs and internal utilities in Python, TypeScript, Rust, or whatever the job needs—delivered with tests, docs, containers and CI from day one.
Security Review & Secure-by-Default Engineering
Review of code and of the AI-enabled tools showing up inside companies without an engineering team behind them. Dependency, secret and supply-chain scanning baked into CI. Cleared and NIST 800-171-literate, without making compliance the product.
Data Analytics & Visualization
Pipelines and dashboards that turn raw data into decisions. Python, SQL and web-based visualization, from a one-off analysis to a scheduled ETL that keeps running after I leave.
Agent Workflow Enablement
Help your team build the way I do: parallel coding-agent sessions, model-per-task selection, review gates, and project memory that survives a context reset. Hands-on, with your codebase.
About
The person behind the code
I'm Matthew Bonham, founder of Bonham Technologies. I'm a software engineer by trade and a systems engineer by choice—most comfortable where technology meets business.
Over the past year the way I build changed. I now run several coding-agent sessions in parallel, choose the model for each task, settle the architecture and specs up front, and spend my time on review and direction rather than keystrokes. The products below were built that way—and they're built to be run that way: multi-agent, multi-model, and deployable without a cloud.
Constrained environments are still home. I came up in defense and federal work, hold an active clearance, and still do the security work—but the product is the software, not the compliance paperwork.
I believe software projects are never truly finished until they're transitioned or decommissioned. That means APIs, documentation, test suites, and extensibility aren't afterthoughts—they're built in from day one.
How I work
One engineer, a fleet of agents
Decide first
Architecture, names, data model and specs get settled in a planning session before a line of code exists. Most steering happens here, not mid-build.
Run in parallel
Several agent sessions at once, each on its own project, with the model chosen per task—frontier models where judgment matters, smaller or local ones where it doesn't.
Review, don't retype
I read the diffs, run the tests and redirect. Every repo ships with typed code, multi-stage CI, dependency and secret scanning, and signed releases.
Keep the record
Decisions, runbooks and session logs live in a durable project record, so no context dies with a chat window and any session can pick up where the last one stopped.
Philosophy
Give me a need and an operating environment, and I will find a way to address that need come hell or high water. Creativity in constrained environments is a specialty.
Featured Work
Products & Projects
ESCAPESUITE
Open-source, privacy-first media creation
A free, MIT-licensed suite of browser-based tools for screen recording and video editing that process entirely on the client. No uploads, no cloud, no accounts—just secure, local processing that works even in airgapped environments.
- ESCAPECRAFT for screen & webcam recording
- ESCAPEARTIST for timeline-based video editing
- WebCodecs-powered encoding & decoding
- Free & open source — single-file offline builds
COGLOSSA
In developmentMulti-agent translation pipeline
Literal and contextual translator agents, a reviewer that reconciles them, a judge that scores error spans, and a round-trip check—each on a different model. Runs against OpenAI-compatible backends, airgapped or hosted, from one engine.
CORELATA
Private demoAnalytic reports from an agent team
A value gate, a writer, a conveyance reviewer, and a compliance auditor that never edits—producing finished ICD 203-style assessments against a standards pack the customer can replace. Unattended from kickoff to delivered report, with every source traceable.
UNVARNISH
In developmentAgentic code review
Built for the AI-enabled tools popping up inside companies without an engineering team behind them, and just as happy reviewing a traditional codebase. Finds what a rushed human reviewer won't, and says so plainly.
WICKMILL
In designAgent orchestration as a world
Characters with a role, a model and tools; buildings where they work; a map you can watch run. Designed for people who have never run an agent, with the power features layered in for those who have.
Sponsors
Who keeps the open-source work going
ESCAPESUITE is free, MIT-licensed, and will stay that way. The people and companies below sponsor Bonham Technologies on GitHub to keep it maintained. The upper tiers are a support retainer—hours with me each month, priority on issues, and a direct line—so a team that relies on the tool has someone to call.
- Open
- Open
- Open
- Be the first
Need an SOW and an invoice instead? Email me and we'll do it that way.
Contact
Let's talk
A tool you need built, an agent workflow you want to stand up, or a system that has to run with the network cable unplugged—I'd enjoy hearing about it. I typically respond within one business day.