Two ways to work with us. Bot work is fast and yours to verify. With an engineer behind the work, you skip reviewing it yourself, you get senior eyes on it, and you have someone to ping when things change.
First $50 free per workspace — no credit cardThe hardest parts of engineering — judgment, ownership, being there later — are the parts current AI agents don't try to do. We added them back.
You don't pick a tier — the bot does. It reads what you're asking for and routes you to whichever path actually fits the work. Here's what each looks like.
Flat prices, not ranges. The bot picks a tier from the request and names the price up front. The bot caps at $80 — anything bigger or judgment-shaped routes to Consulting instead.
| Scope | Flat price | Example |
|---|---|---|
| Q&A · explain · micro-snippet | $0 | "what's the diff between map and forEach?" — no quote, just answer |
| 🪶 Micro · one-line tweak, rename, format, translate | $3 | "rename this var to camelCase", "format this JSON" |
| 🔧 Small · single bug fix · one endpoint · one component | $10 | "fix the off-by-one in lineNumber.tsx", "add input validation to /api/signup" |
| 🛠 Medium · multi-file refactor · write a module · code review | $30 | "refactor the date utils to Temporal", "review this 800-line PR" |
| 🏗 Large · new project scaffold · complete feature · SDK integration | $80 | "scaffold a Next.js admin dashboard with auth + DB" |
| ↗ Beyond Large? | → Consulting | Anything bigger or that needs design judgment routes to a human |
Time booked with an experienced engineer. Every session ends with a written deliverable: a decision doc, an architecture sketch, a code review summary, or a strategy memo. Prep work is done by the bot ahead of the session so the human time is spent on judgment, not background reading.
| Format | Price | What you get |
|---|---|---|
| 📞 Quick Call · 30 min | $200 | Sanity-check a small decision · PR review · "is this the right approach?" with one-paragraph written summary |
| 🎯 Solution Design · 60–90 min | $500 | Architecture for one feature or subsystem · decision rationale · first-step implementation plan |
| 🏗 Design Workshop · half-day (3–4 h) | $1,500 | System-level design or major selection · decision doc with alternatives · phased roadmap · risk register |
| 🤝 Fractional Engagement · monthly | $3,000+ | Embedded advisor · weekly sync · async Slack always-on · custom scope. Email cz2440@columbia.edu to scope. |
When the bot routes you to Consulting (or you ask for it directly), it shows a booking link. You pick the tier, share a few sentences about what you're solving, and our system pages an available engineer. Typical lead time:
The bot classifies during the quote step. ANY of these triggers routes you to Consulting (with-human) instead of Bot Work. Multiple triggers don't stack — it's a binary flip.
deploy, live users, prod DB, config flag rollout, feature toggle. If the user can be affected, a human looks.DB migration, schema change, force push, delete data, uninstall, retract release. Things that hurt if wrong.Every substantial request goes through this. Q&A and tiny edits skip the quote — miniCTO just answers.
Three examples covering Auto, Human, and ambiguous.
src/utils/date/ to use Temporal. About 6 files, similar shape, swap call sites.portal/topup to unlock — once paid, you'll get a booking link.
A single piece of work with a beginning and end. Usually = a Slack thread, or a slash command invocation. If you ask for follow-up changes in the same thread, that's a continuation — no new quote unless the scope materially grew.
So a team of 20 doesn't get $1000 of free agent time by virtue of headcount. The workspace is the billable unit; the free quota matches.
No. Top-up credit rolls over month-to-month indefinitely. If you uninstall, unused credit is refunded.
If real cost ends up under the quoted range, we charge the actual cost, not the quote. If real cost would exceed the quote, the bot pauses mid-work and asks before continuing. We don't surprise you with a bigger bill.
Time with an experienced engineer — currently a small bench, growing. For $200 you get a 30-min Quick Call. For $500 you get a 60–90 min Solution Design session. For $1500, a half-day workshop with a written decision doc. Every tier ends with a written deliverable; the call isn't just chat. The bot does prep ahead of the session so the human time is spent on judgment, not status updates.
Senior ICs and tech leads with 8+ years building production systems at companies you've heard of. The bench is intentionally small — we don't subcontract to people we wouldn't hire ourselves. Profiles available on request before booking.
Yes — just say so in chat. "Book me a consult" or "I'd like an engineer to look at this" routes you to the Consulting flow regardless of what the bot would otherwise quote. Same the other way: if the bot flags Consulting and you want it to just go ahead with Bot Work instead, say "go anyway" — you'll get the bot's output, with no engineer review behind it.
Software estimates are uncertain. A point estimate would either be too conservative (we leave money on the table) or too aggressive (we'd hit the wall mid-work). The band gives you a realistic expectation; we charge the actual cost within the band.
Not yet — usage patterns vary enormously per team, so a flat per-seat number doesn't fit anyone. Per-requirement matches what you actually use. If your team scales to "we want unlimited agent time", email and we'll talk enterprise.
CLAUDE.md teaches the bot this logic per-channel. Triggers, ladders, and quote scripts are all in src/workspaces.js. Update there, then run refresh to push to all channels.