Show HN: Addom – Open-source, telemetry-free AI coding harness and editor
ADDOM is a local-first, open-source Electron desktop workspace for multi-provider AI coding, combining conversation, project-aware tools, an integrated Monaco editor, and delegated agent roles It emphasizes privacy by storing all project records, thread history, credentials, and local memory on-device with no remote telemetry from the developer The application supports configurable agent roles with bounded permissions and isolated write workflows, allowing users to delegate tasks to specialist A
Analysis
TL;DR
- ADDOM is a local-first, open-source Electron desktop workspace for multi-provider AI coding, combining conversation, project-aware tools, an integrated Monaco editor, and delegated agent roles
- It emphasizes privacy by storing all project records, thread history, credentials, and local memory on-device with no remote telemetry from the developer
- The application supports configurable agent roles with bounded permissions and isolated write workflows, allowing users to delegate tasks to specialist AI agents
- Built on Node.js 24 with a React renderer, ADDOM uses a narrow versioned bridge (
window.addombridge) for secure cross-process communication - Currently at version 0.1.0-alpha with preview builds for Windows, macOS, and Linux; released under the MIT License
Why It Matters
ADDOM addresses a growing industry need for local-first, privacy-conscious AI coding tools that don't sacrifice multi-provider flexibility. Its agent delegation model with bounded permissions offers a practical pattern for enterprise and individual developers seeking controlled AI integration into existing workflows without exposing sensitive code to telemetry.
Technical Details
- Architecture: Electron-based desktop app with a three-layer source structure—
src/mainhandles Electron integration, persistence, providers, tools, agents, and credentials;src/preloadexposes a narrow versionedwindow.addombridge;src/renderercontains the React interface, editor, and settings - Multi-provider support: Users can configure cloud API keys, supported account authentications, and local model providers, then select provider and model per thread
- Guarded tooling: Includes guarded filesystem, search, command, and terminal tools with configurable permission modes; agent roles have bounded permissions and isolated write workflows
- Integrated editor: Built on Monaco, providing project source, symbol outline, problems, and reviewable change history alongside the conversation that produced them
- Local-first storage: All project records, thread history, credentials, settings, attachments, and memory are stored on-device; remote model usage still sends prompts and context to the selected provider
Industry Insight
- The local-first, no-telemetry approach positions ADDOM well for regulated industries and security-conscious teams adopting AI coding assistants, filling a gap between cloud-only tools and fully offline solutions
- The bounded-permission agent delegation model reflects an emerging best practice: giving AI agents scoped authority rather than unrestricted access, which reduces risk in collaborative and production environments
- As AI coding tools mature, expect increased demand for transparent, auditable workflows with explicit change review—ADDOM's artifact history and reviewable changes pattern may influence future tool design
Disclaimer: The above content is generated by AI and is for reference only.