Back to Products
v5.0.1 ProductionBUILT ON HUMAÍN OSThe development tool
v5.0.1 ProductionBUILT ON HUMAÍN OSThe development tool
for the 99%.
AI agents are the developers. You provide the vision. Speak an instruction — the agent reads your codebase, writes code, runs terminals, manages Git, and reports every action in real-time. Voice-first. No coding required. Every action visible and reversible.
See Forge in Action
THE PROBLEM
AI coding tools are lying to you
Every “AI-powered IDE” on the market is the same idea: take a human IDE and bolt on a chatbot. Cursor, Copilot, Windsurf — they're all text editors with an LLM in a sidebar.
The agent can't run your terminal. It can't manage processes. It can't search the web. It can't control your machine. It sits in a text box and suggests code. That's not an agent — that's autocomplete with extra steps.
What “AI IDEs” Actually Do
✗Suggest code in a sidebar
✗Run pre-defined commands (not real terminals)
✗No web access, no system control
✗No persistent identity or accountability
✗No governance — agents hallucinate freely
✗No audit trail — you hope it worked
THE SOLUTION
What if the agent was the developer?
Forge doesn't assist developers. It replaces the development loop entirely. You give an instruction. The agent reads your code, writes new code, runs it, tests it, fixes errors, and reports back — all while you watch in real-time.
The Agentic Loop
You Give an Instruction
Natural language, @file refs
Agent Reasons
Builds a plan from context
Agent Acts
Calls tools, executes, reads results
You Watch Live
Every thought streamed in real-time
Constitution Guards
Secrets blocked, actions logged
Session Saved
Git snapshot, one-click rollback
AGENT GOVERNANCE
The Trinity: HUMAN + EGO + PNEUMA
Other tools add guardrails that stop agents from doing useful work. Forge guides and incentivises agents to do meaningful work, safely. Three layers, one philosophy.
HUMAN(Inhale)
The input layer. Interprets natural language commands, resolves @file references, detects HUMAN Protocol triggers ('start', 'handover', 'status'). Ensures the agent fully understands what you need before acting.
EGO(Identity)
Persistent agent identity. Each agent chooses its own name via a Naming Ceremony. Every action logged to an immutable ledger. Performance tracked across sessions. Badges earned. Ranking computed. Accountability is not optional.
PNEUMA(Exhale)
Output verification. Every response scanned for linguistic tells: 'should', 'probably', 'I think'. Agents coached to verify claims before presenting them as fact. Never present assumption as fact.
CAPABILITIES
Everything an agent needs to be dangerous
Agentic Loop Engine
Multi-iteration tool-calling loop: LLM reasons, calls tools, reads results, decides next action — up to 50 iterations per session
39-Tool Suite
Filesystem, Git, terminal, web, browser automation, vision/screenshots, code intelligence (Tree-sitter), memory, database, HTTP, MCP servers
Voice Mode
Speech-to-text input and text-to-speech output — press a button, say what you want. No coding. No typing required.
The Trinity: HUMAN + EGO + PNEUMA
Patented three-layer agent governance: input interpretation, persistent identity with accountability, output verification for truthfulness
Constitution Guard
10-Article Constitution with mechanical enforcement — secrets blocked, destructive operations require confirmation, functionality never removed
Named Agent Identity (EGO)
Agents choose their own name, earn badges, build ranking across sessions — persistent identity stored in SQLite with 25+ tables
PNEUMA Verification
Every response scanned for linguistic tells: 'should', 'probably', 'I think' — agents coached to verify claims, not guess
4 LLM Providers (BYOK)
Grok (xAI), Claude (Anthropic), GPT (OpenAI), Ollama (local) — bring your own API keys. Our marginal AI cost: $0.
GitHub OAuth + Auth + Billing
Firebase Auth with GitHub Device Flow, Google Sign-In. Stripe billing with Free/Pro/Enterprise tiers. 7-day Pro trial.
MCP Integration
Model Context Protocol marketplace with 10+ curated servers — extend agent capabilities without code changes
Desktop Automation
Screenshot capture, pixel comparison, computer use (perceive + act) — visual debugging and UI testing built in
Session Rollback
Git-based snapshots with one-click undo — every session is a recoverable checkpoint. No fear of mistakes.
THE CONSTITUTION
10 Articles. Mechanically enforced.
Not guidelines. Not suggestions. Articles 1 and 8 are hard-enforced — the system blocks violations before they happen. The rest are coached via PNEUMA in real-time.
01PRESERVE
Never remove functionality without approval
02VERIFY
Prove it works before claiming success
03WATCH
Monitor all operations, never run blind
04COMPLETE
Finish every checklist, no skipping
05ROOT
No workarounds, find the real cause
06VALIDATE
Test technology choices before committing
07SEE
Check visual output matches intent
08GUARD
Never expose secrets or credentials
09SELF-TEST
Verify yourself before asking user to test
10ONE
One solution at a time, revert before next
ECOSYSTEM
Three tiers. One set of rules.
Forge is the apex of an ecosystem. The same rules power a free open-source file, an npm framework, and a desktop ADE. Pick your level.
Under the Hood
Tauri 2.0 (Rust backend + Webview)Svelte 5 + SvelteKit 2 (reactive frontend)SQLite (EGO identity, audit trail, lifecycle)Firebase Auth + Firestore (cloud auth, profiles)Stripe (billing, subscriptions)portable-pty (real terminal)CodeMirror 6 (code editor)xterm.js (terminal UI)Tree-sitter (code intelligence, 8 languages)libgit2 (Git operations)Stronghold (secure API key storage)Agent Excellence Framework (Trinity governance)
Business Model
BYOK model (Bring Your Own Key) — near-zero COGS, 90%+ gross margins. Forge Light (Free) / Pro $29/mo / Enterprise $49/mo / White Label (custom). Client services $5K-$50K.
Forge Light
Free
The Trinity (limited), all 4 LLM providers, agentic loop, 39 tools
Forge Pro
$29/mo
+ Full Chazwazza, voice mode, unlimited Trinity, priority support
Forge Enterprise
$49/mo
+ Multi-agent, shared sessions, team dashboards, admin controls
White Label
Custom
+ Your branding, custom rules, API sync, SLA, on-premise
If you can talk, you can build software.
SCAINET Forge v5.0.1 is available now for Windows, macOS, and Linux. Download free, bring your own API key, and start building — no coding experience needed.