Product 01 — Deep Agent CLI

Diana.
The world's A.I.,
one terminal.

Diana is a Deep Agent CLI built for programming. It doesn't lock you into a single A.I. provider — it gives you all of them, orchestrated from a single command line, so you always use the best model for the job.

The basics

What is a
Deep Agent?

You know chatbots — you ask, they answer. A Deep Agent goes further. It doesn't just reply. It thinks, plans, executes, and iterates — autonomously. It reads your code, understands your architecture, runs commands, fixes errors, and delivers results. Not a suggestion. A solution.

Think of it this way: a chatbot is a consultant who gives advice. A Deep Agent is an engineer who sits down, opens the codebase, and does the work.

01

Understands context

It reads your entire project — files, structure, dependencies — before making a single decision.

02

Plans before acting

It breaks your request into steps, considers trade-offs, and architects a solution before writing any code.

03

Executes autonomously

It writes code, creates files, runs tests, and fixes its own mistakes — without you babysitting.

04

Iterates on failure

When something breaks, it doesn't stop. It reads the error, adjusts its approach, and tries again.

LLM Agnostic

Every A.I. has
a superpower.
Diana uses them all.

Not every model is the best at everything. Claude is exceptional at code. Google generates stunning images. GPT excels at structured planning. Why choose one when you can have all of them?

Diana is LLM agnostic — it routes each task to the model that does it best. You don't configure anything. You just work. Diana handles the orchestration transparently, behind the scenes.

Claude
Code & Architecture
Deep code understanding, refactoring, debugging, and architectural reasoning. When Diana needs to write or reason about code, Claude is the engine.
Google Gemini
Image Generation
Need mockups, diagrams, or visual assets? Diana routes image tasks to Google's models, producing results directly in your workflow.
GPT
Planning & Strategy
Work plans, task breakdowns, project roadmaps. When you need structured thinking and project orchestration, GPT handles the strategy layer.
+ Any model
Open Architecture
Diana's architecture is provider-agnostic. As new models emerge and excel at new tasks, Diana integrates them — your workflow only gets better.
$ diana "Refactor the auth module and generate a diagram of the new flow"
Routing: code refactor → Claude
✓ auth/middleware.ts — rewritten with session validation
✓ auth/handlers.ts — split into command/query handlers
✓ auth/tests/middleware.test.ts — 12 tests passing
Routing: diagram generation → Google Gemini
✓ docs/auth-flow.png — architecture diagram generated
2 models. 1 command. You didn't choose — Diana did.
Philosophy

Zero Lock-in.
Maximum
productivity.

Most A.I. tools tie you to one provider. If that provider raises prices, degrades quality, or shuts down a model — you're stuck. Diana eliminates that risk entirely.

Diana is the harness that makes the world's A.I. available for programming, regardless of provider. Switch models with zero friction. Your workflow stays the same. Your productivity never depends on a single company's roadmap.

0
Vendor Lock-in
N
Models available
1
CLI to rule them all
MCP Compatible

Connect to
any tool.
Any service.

Diana is fully compatible with MCP — the Model Context Protocol.

Think of MCP as a universal adapter. Just like USB-C lets you plug any device into any port, MCP lets A.I. agents plug into any tool — your database, your API, GitHub, Slack, Jira, your internal systems. One standard. Every connection.

Without MCP, every integration is custom-built. With MCP, you connect once and every A.I. model Diana uses can access that tool. It's the difference between wiring every appliance directly to the power plant — and just having a wall outlet.

What MCP does

It creates a shared language between A.I. agents and external tools. The agent says "read this file" or "query this database" — MCP translates that into the right API call, no custom code required.

Standard Protocol

Why it matters

Without MCP, your agent is isolated — it can only talk and think. With MCP, it can act: read databases, push code, send messages, trigger deploys. MCP turns A.I. from a brain into hands.

Actionable A.I.

Diana + MCP

Diana speaks MCP natively. Connect any MCP server — community-built or your own — and Diana's agents can use those tools immediately. Your infrastructure, your rules, Diana's intelligence.

Plug & Play
$ diana mcp add github
✓ MCP server "github" connected
$ diana "Create a PR with the refactored auth module"
Using MCP: github → create pull request
✓ PR #247 created — "Refactor auth module with session validation"
Diana wrote the code. MCP pushed it. You sipped your coffee.
Get Started

The best A.I. for every task.
One CLI.

Diana is the harness that makes the world's A.I. available for programming. No lock-in. No compromise. Just the right model, every time.