Craig is a local AI software engineer for solo builders, agencies, and small teams. He reads the repo, plans the task, edits files, runs checks, and shows his work.
Craig is a product character, not a fake employee. The personality is simple: practical, direct, protective of the codebase, and allergic to vague tickets. Craig reads before he writes.
Craig should be proven through short videos: prompt, inspection, edits, commands, result. The viewer should see the work happen.
Demo 01: Build a feature
Prompt:
Add PDF export to the report page.
Craig:
- maps routes
- finds report component
- edits export action
- adds button state
- runs build
- reports diff
Proof:
Changed files + build result.This is the strongest hiring proof: Craig turns a rough product request into inspectable code.
Demo 02: Fix a bug
Prompt:
The contact search breaks on empty filters.
Craig:
- reproduces behavior
- traces state path
- patches condition
- adds narrow test
- runs checks
Proof:
Bug explanation + passing test.Debugging demos show judgment, not just generation. That is what makes Craig credible.
The workflow is intentionally legible. Craig should never feel like a black box making mystery changes to a codebase.
Inspects project structure, key files, package scripts, and existing patterns before touching code.
Turns the request into a small implementation path with clear files and checks.
Makes targeted changes instead of rewriting everything. The diff matters.
Executes build, lint, tests, or project commands where possible and streams progress.
Explains what changed, what passed, what failed, and what still needs review.
Craig connects naturally to the rest of the ecosystem. Each tool gives the coding agent one missing production skill.
Helps Craig understand code structure, source chunks, and where changes belong.
Gives Craig precise search-based edits and diff-first code changes.
Checks generated code for common security, bug, and style issues.
Lets Craig remember decisions, bugs, and implementation notes across sessions.
Wraps tool calls with retries, timeouts, JSON repair, and failure metadata.
Checks URLs, JSON, dates, and structured outputs before downstream use.
Craig is built for the moments when a small team needs software motion without pretending the agent should replace engineering judgment.
Turn a rough product idea into a working branch or demo surface.
Trace the issue, patch the behavior, and run a narrow verification path.
Fill in missing coverage, usage examples, setup notes, or README improvements.
Give agencies and solo builders fast product surfaces they can inspect and refine.
Make small, visible improvements without losing the current architecture.
Help new contributors understand project structure and implementation paths.
Genesis is the marketplace layer: Craig discovers new capabilities, installs them, and uses them in future workflows. The product story is simple: Craig starts as a coding agent, then grows through skills.
The marketplace page should show Craig's installed skills, recent builds, verified capabilities, and available upgrades.
The X account should be the daily proof stream and lightweight advertising channel: demos, diffs, website audits, business before/afters, CRM flows, and short notes about what Craig is building.
Keep it simple: one task, one short clip or screenshot, one result. Craig's voice should be practical, direct, and focused on showing the work behind both the project and Craig the Builder.
Today Craig added CSV import.
Read: contacts page + data layer
Changed: 3 files
Checked: build passed
Next: validation edge cases
Craig reads before he writes.Digital readiness audit:
Business: local contractor
Score: 42/100
Missing: quote form, service pages, CRM
Craig the Builder:
site + lead capture + follow-up.That is the brand: not magic, not fake autonomy, not replacing developers. A practical local coding agent that shows its work.