Screenshots 1
Generative IDE
Summary
Most AI code editors will happily rewrite half your codebase the moment you tab-complete the wrong suggestion — no confirmation, no diff, no way back. GIDE puts a gate on every edit before it touches a single line.
GIDE is a desktop AI code editor built for teams that cannot or will not send source code to a third-party API. It runs local models fully offline, supports cloud models (Claude, GPT, Gemini) via your own API key, and gates every proposed change through an explicit accept/reject step before writing to disk. The agent reads files, runs tool calls, and shows diffs — all inside the editor. The gate is not optional decoration: the vendor states plan confirmation before edits is a core design constraint. Where it breaks: teams expecting background auto-apply will fight the confirmation model, and the toolchain is new enough that community-sourced workarounds are sparse.
Bottom line: The right pick for an air-gapped team that needs local inference with zero API leakage and is willing to approve every change; a friction point for teams that want a 'just fix it' agent that runs unattended.
Pricing Plans
Subscription- Price
- $19.99/mo
Developer
Ornith 9B bundled, unlimited local inference, plan-gate, refactoring engine
- 30-day free trial
- Email support
Business
Everything in Developer plus BYOK cloud models, MCP client, context memory, GIDE Wiki
- Priority email support
Enterprise
Everything in Business plus air-gapped deployment, SSO/SAML, custom compliance
- Dedicated onboarding
- 1-day support SLA
View full pricing on generativeide.com →
Pricing may have changed since last verified. Check the official site for current plans.
Community Performance Report Card
No community ratings yet. Be the first to rate this tool!
Pros
Sign in to edit- Runs local models fully offline at zero token cost, so teams in air-gapped or regulated environments get AI assistance without any source code touching an external API.
- Write gate requires explicit accept or reject before any file is modified, so a runaway agent cannot silently overwrite production logic while you are looking at another window.
- Model-agnostic design lets you swap between local inference and cloud models (Claude, GPT, Gemini via your own API key) without changing your workflow, so a cost spike or model-quality issue is not a migration project.
- Semantic repo retrieval reads only the files the task actually requires, so context windows stay focused and the agent is not padding every prompt with irrelevant code.
- Self-hosted option and available API mean teams can deploy a consistent toolchain across developer machines without depending on a vendor-controlled cloud service going down or changing pricing.
Cons
Sign in to edit- The write gate confirmation model adds a manual step to every discrete edit. Teams running large-scale mechanical refactors — renaming a symbol across 200 files, for example — will click through approval prompts repeatedly or look for a way to batch-accept, which the current docs do not clearly describe. Teams that need unattended bulk changes will evaluate competitors with auto-apply modes instead.
- There is no perpetual free tier. The trial ends, and teams mid-integration who have not yet hit a real production workflow will be asked to commit before they have enough signal. Teams that need a longer evaluation runway before budget approval will stall here.
- MCP integration for external tools is listed as a feature, but the product page provides no detail on which MCP servers are supported or what the setup path looks like. Teams expecting plug-and-play connections to their existing toolchain will spend time on undocumented integration work before they get value from that layer.
About
- Platforms
- macOS, Windows, Linux
- API Available
- Yes
- Self-Hosted
- Yes
- Last Updated
- 2026-09-08T22:32:24.609Z
Best For
Who it's for
- Privacy-conscious developers
- Air-gapped or regulated teams
- Users preferring local inference
What it does well
- Offline AI coding in secure environments
- Multi-file refactoring with previews
- Project memory and architecture visualization
- Connecting to external tools via MCP
Integrations
Add notes, reviews, and benchmarks so the next visitor gets a clearer picture.
Spotted incorrect or missing data? Join our community of contributors.
Sign Up to ContributeFrequently Asked Questions
- Is Generative IDE free?
- Generative IDE is a paid tool ($19.99/mo). A 30-day free trial is available.
- Is Generative IDE open source?
- No — Generative IDE is a closed-source tool. Source code is not publicly available.
- Does Generative IDE have an API?
- Yes. Generative IDE exposes a developer API. See the official documentation at https://generativeide.com for details.
- Can I self-host Generative IDE?
- Yes. Generative IDE supports self-hosting on your own infrastructure.
- What platforms does Generative IDE support?
- Generative IDE is available on: macOS, Windows, Linux.
Best Generative IDE alternatives →
Curated lists that include this category
GIDE is a model-agnostic AI code editor that runs on your machine. The core workflow is a chat-driven agent loop: you describe a task, the agent reads the files it needs, proposes a diff, and waits for you to accept or reject before writing anything. The vendor describes this as a ‘write gate’ — it is on by default and the demo sessions on the product page show the approval prompt appearing between every discrete edit. Local models run at zero token cost; cloud models connect via API keys you supply yourself.
The differentiating feature is the combination of offline-first architecture and a plan-confirmation system. The vendor states GIDE is air-gap ready for regulated industries, which means the network dependency is optional rather than baked in. On local models, there are no per-token charges and no data leaves the machine. The agent uses semantic retrieval across the repo — glob patterns, file outlines, targeted reads — rather than loading the entire project into a context window at once, which is how it handles larger codebases without ballooning prompt size.
GIDE fits teams in security-sensitive environments: regulated industries, defense-adjacent work, or anyone whose legal team has flagged code-to-cloud data flow. It also fits developers who want to stay in control of what changes — the approval gate is a feature, not a workaround. Where it breaks: the confirmation-first model is slow for bulk mechanical changes where you trust the agent completely. There is no perpetual free tier, so teams evaluating it hit the trial cutoff before a full production integration cycle. MCP (Model Context Protocol) connectivity is described for external tool integrations, but teams needing deep CI/CD pipeline hooks will be building that glue themselves.
The editor ships with a CLI installer for macOS, Linux, and Windows. Version 2.3.6 is documented on the product page. The API is available, and self-hosting is supported — relevant for teams that need to standardize tooling across multiple developer machines without relying on a vendor-managed cloud instance.