Open-Source Chrome extension for AI-powered web automation. Run multi-agent workflows using your own LLM API key. Alternative to OpenAI Operator.
13.4k
Stars
1.4k
Forks
69
Open issues
25
Contributors
AI Analysis
Nanobrowser is an open-source Chrome extension for AI-powered web automation that runs locally in your browser using your own LLM API keys. It provides a free alternative to OpenAI Operator with support for multiple LLM providers and a multi-agent system architecture. It serves developers, automation engineers, and organizations seeking cost-effective, privacy-respecting browser automation without cloud dependencies.
Inferred from signals mentioned in the README (tests, CI, type safety) — not a review of the actual code.
AI's overall editorial judgment — not an average of the bars above, can weigh other factors too.
Open-source Chrome extension brings multi-agent AI web automation to users who supply their own LLM keys
Nanobrowser is a Chrome/Edge extension that lets users automate web tasks through a multi-agent AI system (Planner + Navigator roles) using their own LLM API keys from providers like OpenAI, Anthropic, Gemini, or local Ollama. It targets privacy-conscious developers, power users, and AI enthusiasts who want browser automation without paying for hosted services like OpenAI Operator. The extension runs entirely client-side, keeping credentials local. With 13K+ stars and a Chrome Web Store listing, it has genuine public visibility, though depth of production adoption beyond hobbyist use remains unclear.
Created at the very end of 2024 (December 31), Nanobrowser emerged directly in response to OpenAI Operator's launch, positioning itself as a free, self-hosted alternative. It is less than 18 months old as of the evaluation date.
The project captured early attention by timing its launch with the OpenAI Operator announcement, riding the hype wave around agentic browser automation. Most star growth likely came in a burst during early 2025. However, with only 30 stars gained in the last 7 days as of mid-2026, growth has clearly plateaued well below the pace of category leaders like browser-use (100K+ stars), suggesting the initial viral moment has passed.
A Chrome Web Store listing with a published extension ID is a concrete distribution signal. The Discord community and Twitter/X presence suggest an active user base beyond pure star-gazers. However, no verified download counts, enterprise use cases, or third-party integrations are cited in the available metadata. Adoption appears primarily among hobbyists and developers experimenting with browser agents rather than verified production deployments.
Appears to implement a multi-agent architecture with at least two specialized agents (Planner and Navigator) communicating via a Chrome extension side panel. Likely uses Chrome Extension Manifest V3 APIs for browser control. Built in TypeScript with pnpm as the package manager and Node.js v22+ as the runtime. The separation of Planner and Navigator roles suggests a task decomposition pattern where one agent reasons about strategy and another executes DOM interactions.
not documented in README
Last push was November 24, 2025 — approximately 7 months before the evaluation date of June 2026. This is a notable gap and raises questions about whether active development has paused or slowed significantly. The project is not stagnant by historical standards, but the absence of recent commits in 2026 is a yellow flag for a project in a fast-moving space.
ADOPT IF: you want free, privacy-preserving browser automation via a Chrome extension, are comfortable supplying and managing your own LLM API keys, and need a working solution today rather than a production-hardened one. AVOID IF: you need reliable, maintained software for business-critical automation — the 7-month gap since the last push raises real sustainability questions for a project in an active competitive space. MONITOR IF: you are tracking the browser-agent extension category and want to see whether development resumes or the project is superseded by a more actively maintained fork or successor.
Independent dimensions
Mainstream potential
3/10
Technical importance
6/10
Adoption evidence
3/10
- Development appears to have stalled since November 2025 with no public commits in 2026, which is a significant concern in a category evolving rapidly.
- Entirely dependent on Chrome/Edge extension APIs; any breaking change in Manifest V3 or Chrome's extension platform could require substantial rework.
- The multi-agent approach requires users to pay for multiple LLM calls per task, potentially making complex workflows expensive despite the 'free' positioning.
- Category has fast-moving competition — browser-use and similar Python-based tools may absorb the developer audience, leaving Nanobrowser's growth ceiling low.
- No documented test coverage or CI/CD signals in the README means code quality and regression safety are uncertain for users who need dependable automation.
Nanobrowser is likely to remain a useful hobbyist tool but may struggle to grow meaningfully in 2026 given apparent development slowdown and intensifying competition from more actively maintained alternatives.
Newsletter
Get analyses like this every Monday
Free weekly digest of the most interesting open-source discoveries.
Languages
Information
- Website
- https://nanobrowser.ai
- Language
- TypeScript
- License
- Apache-2.0
- Last updated
- 8mo ago
- Created
- 19mo ago
- Analyzed with
- anthropic/claude-haiku-4-5
Stars over time
Contributors over time
Top 100 contributors only — repos with more will plateau at 100.
Open issues
[Docs]: Gemma 4
[Bug]: removeThinkTags step 2 wipes legitimate content after a stray </think>
Question about featuring Nanobrowser as a browser agent app
[Feature]: Why doesn't it have a behavior similar to Claude? It should work with the TAB open.
Top contributors
Similar repos
ntegrals/openbrowser
Open Browser is a TypeScript framework that enables AI agents to autonomously...
browser-use/browser-use
Browser Use is a Python library that enables AI agents to control a real web...
| Repository | Stars | Week Δ | Language | Score | Updated |
|---|---|---|---|---|---|
|
nanobrowser/nanobrowser
this
|
13.4k | +37 | TypeScript | 7/10 | 8mo ago |
|
ntegrals/openbrowser
|
9.5k | — | TypeScript | 8/10 | 3mo ago |
|
browser-use/browser-use
|
104.1k | — | Python | 9/10 | 4h ago |
 hyperbrowserai/HyperAgent
|
1.5k | — | TypeScript | 7/10 | 2mo ago |
 HKUDS/nanobot
|
45.2k | — | Python | 7/10 | -0 min ago |
 Nano-Collective/nanocoder
|
2.2k | — | TypeScript | 8/10 | 10h ago |
Python-based library with 100K+ stars that controls browsers via Playwright/Selenium from scripts rather than a browser extension. Far more popular and actively developed; targets developers who want to integrate browser automation into backend workflows rather than end-users wanting a sidebar interface.
The web-UI wrapper for browser-use, offering a similar chat-driven browser automation experience but as a standalone web app. Has more stars (16K) and benefits from the browser-use ecosystem; requires running a local server unlike Nanobrowser's extension model.
A comparable Chrome extension approach with 9.4K stars. Likely overlapping target audience. Nanobrowser edges it in star count but the two appear to serve the same niche; differentiation between them is not fully clear from available metadata.
37K stars and written in Rust, suggesting a performance-focused or infrastructure-level approach. Likely targets a different developer profile (systems/backend) rather than end-users installing a Chrome extension. Not a direct substitute for Nanobrowser's use case.
The commercial product Nanobrowser explicitly positions against. Operator requires a $200/month ChatGPT Pro subscription and runs in a cloud-controlled browser. Nanobrowser trades Operator's polish and reliability for cost savings and local privacy, which is a meaningful tradeoff for cost-sensitive or privacy-sensitive users.
