Download BobCode01Models
Every frontier model,
picked per step
Claude, GPT, Gemini, DeepSeek and Qwen behind one plan. The smartest one runs the step that needs it; a fast cheap one runs the rest.
EVERY MODEL
Claude, GPT,
Gemini, DeepSeek
Gemini, DeepSeek
one plan
EVERY RUN
Watch it, pause it,
roll it back
roll it back
IT DELIVERS
Files you can
actually open
actually open
WDOC
PPPT
XXLS
ONE HARNESS · EVERY MODEL · FOUR PLATFORMS
Every job
Every job
finishes itself
Claude, GPT, Gemini and the rest behind one agent harness that plans, runs and finishes the job — connected to the 100+ tools you already work in. Free to start.
Good morning.
What should it build for you?
What should it build for you?
Read the request
done
Map the repository
done
Found 34 files across 4 packages. Planning the change before editing.
Plan the change
approved
Write the code
done
Run the tests
fail
Two tests failed on the idempotency key. Retrying with a narrower fix.
Fix what failed
done
Deploy the preview
running
Runs on every frontier model
SIX CORE CAPABILITIES
Six things it does,
Six things it does,
from the first run
Compose them however the work actually runs.
Scroll sideways to browse →
Every frontier model,
02Delegation
It commands
Claude Code and Codex
It drives Anthropic’s official Claude Agent SDK and OpenAI’s Codex as subagents in your workspace. Neither of them can do that to BobCode.
BobCode
Claude Code
Codex
Claude Code
Codex
03Connectors
100+ connectors
into your own systems
Notion, ChatGPT, Gmail, Sheets, Slack, GitHub, Stripe, HubSpot, Salesforce. It works inside them with your credentials, not beside them.
04Harness
It recovers
on its own
A failed tool call is a branch, not an ending. It retries, backtracks and keeps going for hours after you have closed the laptop.
Run tests
2 failedretry
Fix applied
Tests green
05Console
Every step
on screen
A live console instead of a spinner: which step is running, on which model, touching which system. Roll a run back and take the other fork.
Read requestMap repoPlan changeWrite codeDeploy
06Local
It runs on
your machine
Local-first by default, with an OS-enforced sandbox around file effects. Your repo and your customer data never have to leave to get the work done.
nothing leaves
It does not replace your tools
It commands Claude Code and Codex.
BobCode drives Anthropic’s official Claude Agent SDK and OpenAI’s official Codex as subagents, inside the same workspace, and folds their answers back into the run. Neither of them can do that to BobCode.
BobCodeplans the job, owns the run
Claude Codeofficial Claude Agent SDK
SUBAGENT
Codexofficial codex app-server
SUBAGENT
Verified in the source: subagent-claude-code and subagent-codex.
FOR EVERY ROLE
Each job,
Each job,
its own set
Built around what a role actually repeats.
01
Build a working app
A working internal tool, deployed, in an afternoon.
02
Research and decide
A twenty-source market brief with the sources still attached.
03
Turn mess into a deck
Six months of exports → one board deck, on brand.
04
Work inside a real codebase
The scary refactor, done, with the tests still green.
05
Automate the recurring job
The Monday morning routine, running without you.
06
Design, image and video
Twelve on-brand variants before the meeting starts.
And then you watch it work
Not a spinner. A console.
Every step appears as it happens — which model is running it, which system it is touching, and where it needs you. Pause it, correct it, or roll the whole run back and take the other fork.
1Understand
Slack
Read requestBob AI
Read requestBob AI
GitHub
Map repoBob AI
Map repoBob AI
Plan changeYou approve
2Build
Claude
Write codeOpus
Write codeOpus
DeepSeek
Run testsDeepSeek
Run testsDeepSeek
Fix failuresBob AI
3Ship
Vercel
Deploy previewBob AI
Deploy previewBob AI
Slack
Post linkBob AI
Post linkBob AI
Free
to start — every model included
100+
connectors
Local-first
runs on your machine
Why BobCode
Four reasons teams switch.
Not features. What changes on the day you start using it.
What it costs you
One subscription, not four.
Paying Claude, ChatGPT, Gemini and a coding agent separately is roughly $80 a month for four accounts that cannot see each other’s work. BobCode is one plan from $20 that carries all of them, and they share the same run.
See what it would cost you →Free
to start · $20 a month for every frontier model, not one
- Free to start — every model, no card.
- $20 a month for one person; $25 per user for a team.
- Pro at $100 and Max at $200 when the work runs for hours.
What runs your work
Never be stuck on last month’s best model.
The leaderboard changes every few weeks. If your tool is one vendor’s wrapper, every change is a migration. BobCode routes each step to whichever model is strongest at it — and the day a better one ships, you already have it.
See the model line-up →All of them
frontier models, one subscription, one interface
- Claude, GPT-class, Gemini, DeepSeek, Qwen and local open-weight models side by side.
- Automatic routing picks the cheapest model that can actually do the step.
- Switch model mid-task without losing the thread — the context comes with it.
How far it gets
It finishes. Not just starts.
Most agents are impressive for ninety seconds and then stall on a failed tool call. The harness is the part that plans, retries, backtracks and keeps going — and it is the part we have spent the most time on.
How the harness works →Hours
of unattended work per run, not minutes
- Plans before it acts, so you can correct the plan instead of the mess.
- Recovers from failed tool calls on its own instead of handing you the traceback.
- Runs in parallel across sub-tasks and merges the result.
What it feels like
Watch it work. Take the wheel any time.
A chat log is not visibility. BobCode shows every step as it happens, on a console you can read at a glance — pause it, correct it, or roll the whole run back and try a different way.
Get BobCode →4 devices
macOS, Windows, iOS, Android — plus terminal and web
- A live console instead of a spinner — you can see which step is running and why.
- Roll back any run and re-take the fork you wish it had taken.
- Start on your laptop, check it from your phone, finish in the terminal.
What it replaces
Same work, two ways
What one person pays to keep Claude, ChatGPT, Gemini and a coding agent on hand.
Needs inputcompetitor list prices are public rates as of Aug 2026; re-check before this goes in a deck.
One subscription instead of fourand the models can finally hand work to each other inside a single job.
Connectors
It connects to 100+ of the tools you already run
Notion, ChatGPT, the whole Google suite, Slack, your CRM, your repo, your billing — it reads and writes in them the way a colleague would.
NotionChatGPT
Google DriveGmailGoogle Sheets
Google CalendarSlackGitHubFigmaStripe
LinearHubSpot
Salesforce
Shopify
Airtable
Jira
WhatsApp
ZoomFAQ
Questions
What is BobCode, in one sentence?
An AI agent that does whole jobs — build, research, refactor, report — using whichever frontier model is best for each step, on your desktop, phone, terminal or browser.
How does one plan cover every model?
There is a free plan with every model in it, then Personal at $20 a month, Teams at $25 per user, Pro at $100 and Max at $200. All of them carry every frontier model. Most steps in a job do not need a flagship, so the router sends them to a fast cheap model and saves the expensive one for the steps that actually need it — which is what makes a single plan cover what four subscriptions used to.
Which models can I use?
Claude, GPT-class, Gemini, DeepSeek, Qwen and local open-weight models, plus image, speech and video models. You can let BobCode pick per step, or pin a step to a specific model.
Can I use my own API key?
Yes, and it costs you nothing extra. Bring your key, pay the provider directly, and we take no cut. The full harness is still yours.
Where does my data go?
BobCode is local-first: runs happen on your machine and stay there unless you send them somewhere. Enterprise deployments can run entirely inside your own network.
What happens to my work if I stop paying?
It is already on your machine, because that is where it was made. Bring-your-own-key users keep working against their own provider account with no change at all.