Capabilities

Everything Claude Code does.
With every model.

Open the lid. The whole AI universe — Anthropic, OpenAI, Google, and your own local Idaus — becomes one calm, fast workspace on your machine.

8+
Agent backends
20+
Built-in skills
4
Model families
1
Calm interface
01 — CODE

It writes the diff. You ship the PR.

Refactor a function, restructure a service, ship a feature end-to-end. Gods Eye reads your repo, reasons about your stack, edits files in place, runs the tests, opens the pull request. Not a chat — a teammate that types.

Reads call graphs Edits in place Runs your tests Opens PRs
you · refactor auth.ts to use the new sessions module
Reading auth.ts · 412 lines
Found 6 call sites. Editing in place…
Running vitest auth/
✓ 24 tests passed in 1.2s
Branch refactor/auth-sessions ready · PR opened.
02 — MODELS

One conversation. Every brain.

Switch from Opus 4.7 to GPT-5 to Idaus mid-sentence. Context follows. Pay-per-token cloud models for the heavy lifts, your private local Idaus for everything else. The model is a knob, not a lock-in.

Anthropic · Opus 4.7 OpenAI · GPT-5 Google · Gemini Ultra Local · Idaus
GPT-5 · Codex
Cloud
Gemini 2.5 Ultra
Cloud
Idaus 0.1.0
Local
03 — TOOLS

It runs your shell. Reads your files. Owns the loop.

Gods Eye doesn't reply with a code snippet — it executes. Spawns shells, navigates your workspace, reads logs, hits APIs, edits files atomically. Every action is shown, approved, undoable. You stay in control. It does the typing.

Bash + zsh + pwsh File read/write HTTP requests YOLO or step-by-step
gods-eye · live tool stream
# model decided to inspect failures
~ $ grep -rn FIXME src/ | wc -l
→ 14
~ $ cat package.json | jq .scripts
{ "test": "vitest", "build": "vite build" }
~ $ bun run test
04 — SKILLS

Drop in a skill. The model just knows.

Make a deck. Fill an Excel. Build a financial model. Write a Word doc. Render a Mermaid chart. The Skills Hub teaches any model new tricks — no fine-tuning, no reprompting. Just plug, ask, deliver.

PowerPointExcelWord Financial modelsMermaidCustom
📊Dashboard
📑PPT
📈Excel
📝Word
💰Finance
🌳Mermaid
🎓Academic
🚀Pitch
Custom
A first in AI · Team feature

The end of prompting.

You set the destination.
It writes every step.

For thirty years the deal was simple: you type a prompt, the AI replies, you read it, you type the next prompt. Forever. The work of using AI was the work of writing prompts.

Open a Team in Gods Eye and that loop ends. You give the team a goal — "refactor checkout to use the new payments module and ship a PR" — and the lead agent reads it, breaks it into steps, and writes the prompts to its teammates by itself. Reviewer reviews. Runner runs the tests. Lead reads the result, decides what's next, writes the next prompt. You watch the loop close itself.

This is the first agent engine that doesn't wait for your next message. It plans, it dispatches, it evaluates, it re-plans. Like a small office of senior engineers who don't need to be told twice.

The first agent engine that writes the next prompt for itself, depending on the goal — shipped inside Gods Eye's Team feature.
Goal-driven Self-prompting lead Auto-handoff Live office view
YOUR GOAL
"Refactor checkout to use the new payments module & ship a PR."
looping
01
Leadplan: read payments/*, find call sites in checkout/*, draft migration.
02
Lead → Writer// auto-generated prompt
"Read auth/session.ts and checkout/index.ts. Replace stripeV1.charge with payments.charge. Keep the existing error contract."
03
Writeredited 6 files · 142 lines changed.
04
Lead → Reviewer// auto-generated prompt
"Review the diff in branch refactor/payments. Flag any place we kept stripeV1 references or broke error shape."
05
Reviewer2 nits, 0 blockers. Approved.
06
Lead → Runner// auto-generated prompt
"Run vitest payments/ and checkout/. Capture failures. If green, open a PR titled 'refactor(payments): migrate checkout'."
07
Runner✓ 247/247 pass · PR #1284 opened.
0 prompts you wrote 7 prompts it wrote · loop closed
05 — TEAMS

A team of agents. One tap to assemble.

Spin up a writer + a reviewer + a runner. Give them roles. Watch them collaborate. Claude drafts the code, Codex reviews it, Idaus runs the tests, Gemini summarises the diff. Orchestration without infrastructure.

Lead + teammatesAuto-handoffLive status
06 — MEMORY

It remembers everything. In plain markdown.

Every conversation, every preference, every "remember that I prefer tabs over spaces" gets written to a local Obsidian-compatible vault on your Mac. Come back tomorrow, next month, after a holiday — Gods Eye walks in already knowing your codebase, your style, your in-progress refactor, the variable name you hate.

And because it's just markdown files in a folder you can open in Obsidian, you can edit the memory by hand. Cross out a wrong fact. Add a constraint. The next session reads the diff and behaves accordingly. No retraining, no "context window", no thread-graveyard.

Plain .md files Open in Obsidian Edit by hand Survives every session
~/GodsEyeVault/projects/checkout.md
---
title: Checkout refactor
tags: [in-progress, payments]
last_touched: 3 weeks ago
---
## Where we left off
- Stripe webhook idempotency
- TODO: dedupe by event_id
- Bitan prefers tabs, not spaces
## Decisions
- Use server-side timestamps
- Soft-delete refunds < 24h
// agent reads this on every new chat
07 — COST

Use your own subscription.
Pay zero on top.

Already have Claude Pro? ChatGPT Plus? Gemini Advanced? Plug them in. Gods Eye uses your existing keys — no markup, no proxy, no "credits", no monthly mystery bill from us. The app is free and open-source. Your AI bill is whatever the model vendors charge you, exactly the same as if you were using their websites.

The math people quietly do in their head: cancel Cursor, Copilot, Cline and the half-dozen point tools, keep the one Pro sub you already pay for, get a single calm interface that's better than all of them combined. Many users report saving $40-200/mo this way.

BYO API key Free + Apache-2.0 No per-token markup No telemetry tax
Monthly bills · before & after
Cursor Pro $20
Copilot Pro $10
Cline subs $15
Claude Pro (kept) $20
Gods Eye $0
Net change −$45 / mo
08 — LOCAL

Your prompts. Stay yours.

No telemetry. No cloud round-trip when you don't need one. Idaus runs on the metal of your Mac. Take Gods Eye on a flight, into a SCIF, behind a corporate firewall. It just works.

Local IdausNo telemetry by defaultOpen source
PROMPT FLOW
👤
💻
☁️
You Local · 0ms Cloud (off)

Open the lid.
Skip the tab graveyard.

One window. Every model. Every skill. Every shell. The work, done.

Download Gods Eye