EvolIDE by Evol Group

AI Delivery IDE for serious software teams

Reduce token waste. Keep code private. Ship better software. Server-custodied keys, gateway-only context ranking, and multi-phase agent delivery — built for professional engineering organisations.

Private beta — invite only. Evol Group teams pilot EvolIDE internally today.
0+

Supported models

cloud · local

0

Agent modes

Safe · Assisted · Auto

0

Delivery phases

Plan → Build → Verify → Ship

Works with

OpenAIAnthropicGeminiOllamaLM StudiovLLMllama.cppMistralGroqTogetherOpenAIAnthropicGeminiOllamaLM StudiovLLMllama.cppMistralGroqTogether
57+
Models in advisor catalog
OpenAI · Anthropic · Gemini · local
17
Launch smoke checks
npm run smoke-all
0
Cloud keys on client
Server-custodied, encrypted at rest
BETA
Release stage
Private — invite only

Built for delivery teams.
Governed by design.

Reduce token waste, keep code private, and improve delivery quality with server-custodied keys, gateway-only context ranking, and a local-first default for everyday engineering work.

Agent-native

Multi-phase agent builds

Large tasks auto-split into resilient phases. Each phase saves incrementally — one stall never kills the whole run.

Plan
Build
60%
Verify
Ship

Privacy

Zero keys on client

Cloud provider keys live only on the gateway, encrypted. Your laptop never holds a secret.

Local-first

Offline-capable IDE

Ollama, LM Studio, vLLM, llama.cpp — full agent loop on-device without an account.

Intelligence

Smart model routing

Task-aware advisor with 57+ models. Wrong-model warnings before you burn tokens.

claude-3.7-sonnet0.94
gpt-4o-mini0.81
llama-3.1-70b0.72
ollama/qwen-coder0.66

Operations

Mission control & background agents

Isolated worktrees, task queues, evidence artifacts, browser-to-code fix loops — enterprise delivery from one cockpit.

agent-23 · refactor apirunning
agent-22 · add stripemerged
agent-21 · upgrade depsverifying
agent-20 · port to next 14building

Metering

Transparent token wallet

Per-tenant usage ledger, included grants, hard stops. Know exactly what cloud costs.

Used 6.4k
Grant 10k
Hard stop 12k

Watch an agent ship.
In phases, with proof.

A looping, deterministic walkthrough of an EvolIDE agent run — plan, build, verify, ship — with token metering, evidence, and a green merge gate at the end. Reflects the real cockpit private-beta teams use today.

evolide › agent · session 2f1aDemo

Phases

  1. 01Planrunning
  2. 02Build
  3. 03Verify
  4. 04Ship

Tokens

0/ 3,200

Progress

0%
0/16 events

Policy

SAFEgateway-keysaudit-on
  • plan
model · claude-3.7-sonnet·p50 187ms·cache 71%·budget $0.24 / $5.00looping demo

From prompt to production
without the leaks

01

Open your repo

EvolIDE indexes locally. Context ranking runs on the private gateway when you need cloud-grade retrieval.

02

Agent with guardrails

Safe, Assisted, or Auto modes. Incremental saves, split-view review, rollback, and wrong-model warnings.

03

Ship with proof

Background agents, browser repro loops, benchmark proof, and PR acceptance scores — delivery evidence built in.

evolide — agent cockpit

Budget

$0.24 / $5.00

Build a Next.js dashboard with auth + Stripe portal
Analyze repo & plan scoperunning
Phase 1 · Scaffold & routes
Phase 2 · Gateway wiring
Phase 3 · Portal polish

Your keys never touch
the client.

Provider keys are custodied on the EvolIDE gateway, encrypted at rest. The desktop holds a token only. Context-ranking IP never ships in the client bundle.

  • Client never calls providersEvery LLM call is mediated by the gateway Brain
  • Token-only sessionProfile, prefs, and active model fetched from gateway
  • Brain-mediated contextRanking and routing run server-side, then proxied to LLMs
  • Usage ledgerEvery cloud call metered and auditable
  • Enterprise on-premPrivate gateway + Brain for air-gapped teams

Three product tenets, written into the code.

Not slogans — every tenet maps to a concrete mechanism in the gateway, Brain, or agent loop.

Reduce token waste

Right model, every time.

The advisor scores 57+ models per task and warns before you burn tokens on the wrong one. Brain-side caching and intent classification trim repeats before they hit a provider.

Keep code private

Server-custodied keys. Token-only client.

Provider keys are encrypted at rest on the gateway. The desktop holds a session token only. Ranking and routing logic runs server-side, never shipped to the laptop.

Improve delivery quality

Agents that ship with proof.

Multi-phase agent runs with incremental saves, isolated worktrees, evidence artifacts, and browser-to-code fix loops. Resume from any stall — no more lost afternoons.

Common questions

How is EvolIDE different from other AI IDEs?

EvolIDE pairs an on-device context index with a server-side Django Context Brain that produces an explainable decision before every agent run: intent, ranked context, model recommendation, and a wrong-model warning. Add central managed cloud keys, three transparent usage counters, and a delivery agent loop with evidence artifacts.

Does my source code leave my machine?

Local indexing stays on-device. Only the candidate chunks the brain needs to make its decision are sent to the gateway. A privacy filter drops .env files, keys, and vendor folders before anything leaves your machine.

Can I use local models only?

Yes. Ollama, LM Studio, and OpenAI-compatible local runners are supported. A one-time account activation unlocks a 6-month free local trial; after that the local meter stays free under fair-use limits.

Where do API keys live?

Managed OpenAI / Anthropic / Gemini keys live on the gateway and are resolved at runtime — never copied per tenant. Enterprise BYOK keys are stored encrypted at rest per tenant. End users never see provider keys.

Is pricing available?

Yes — four tiers: Free (local-first), Pro ($19/seat/mo), Team ($59/seat/mo), and Enterprise (custom). Each tier maps 1:1 to a verification-first agent mode. See /pricing for the full breakdown.

What agent modes are supported?

Safe (approve each patch), Assisted (auto-pause on risky ops), and Auto (background worktrees with evidence and rollback).

Ship faster with EvolIDE

We’re onboarding a small set of pilot teams before public launch. Tell us about your stack and we’ll get back to you.

Request early access

Be first to try EvolIDE — AI Delivery IDE for serious software teams.

No credit card. Audit-ready governance from day one.