Fine-tuned open models · no lock-in

AI that plugs into where you work.

Turborg is an AI platform you connect to the tools you already use — your terminal for exhaustive coding sessions with the turborg/borg CLI, IRC across 17 networks, and more connectors coming. Fine-tuned open models, honest pricing, no lock-in.

curl -fsSL https://turborg.com/install.sh | sh

Start free — 50 activities/day ↗ · Read the docs

turborg — ~/api

$ turborg "add a token-bucket rate limiter to the API and test it" · read_file internal/api/server.go ✓· grep "func .*Handler" ✓· edit_file internal/api/ratelimit.go ✓· bash go test ./internal/api/... ✓ ok internal/api 0.41s Added a 100 req/s token-bucket limiter and a passing test. ▌

One agent, many connectors

Turborg meets you where you already work.

Terminal

live

The turborg/borg CLI REPL — an agent that reads, edits, runs, and verifies, for exhaustive coding sessions.

IRC

live

An always-on bouncer with AI polish, summaries, and agents across 17 networks.

coming soon

New connectors are on the way — one account, one budget, wherever you work.

read · edit · run · verify — real tools, not a chat box. Resumable sessions, plain-text output. ~18 MB static binary. Runs as turborg or borg.

the terminal connector

Exhaustive coding sessions

The turborg/borg REPL opens files, edits them, runs your commands, and compiles its own work before calling a task done — recovering from errors itself instead of handing them back.

auth · meter · bill

Metered, never keyed

Every model call is authenticated and metered through your plan. No provider API key lives on your laptop, and spend stays under a rolling daily budget you control.

one account, one budget

Shared across connectors

The same plan powers your terminal and your IRC bouncer. One login, one daily budget of activities, wherever you put Turborg to work.

Honest about our models

We fine-tune open weights. We don't train from scratch — and we don't pretend to.

Turborg runs on open-weight foundations — DeepSeek and Gemma — tuned by us for the platform. You always know what's under the hood. You spend activities, not opaque credits; different models cost different rates.

ModelBuilt onBest forActivityPlan

Chuppa DeepSeek-V4 Everyday coding — fast, cheap reasoning. The default. 1× every plan

Axiom frontier open weights The hardest problems — deep, multi-step work. premium Pro and up

Floko Gemma Light and fast — quick tasks and chat. ≈1× every plan

An activity is roughly one agent turn. Your plan grants a daily pool; Chuppa spends about one per turn, Axiom more for the heavy lifting.

coming soon

Bring your own model.

Point Turborg at any open-weights model you like. No lock-in — not to our models, not to our pricing. Use ours for the convenience and the shared budget, or bring your own. Your call.

Pricing

Start free. Scale when you're ready.

Free

50 activities / day

Get started

Starter

$10/mo

233 activities / day

See all plans

Compare Free, Starter, Pro and Max →

Put Turborg to work.

Install the CLI and start an exhaustive coding session in your terminal — free, no card.