Banana Peel
Banana Peel is a browser-agent routing layer and OpenAI-compatible API. It lets a customer submit one browser task and uses smart routing to select, and when necessary fall back across, hosted agents, browser infrastructure, extraction APIs, and open-source browser frameworks.
At a Glance
- Developers and engineering teams building browser-agent applications
- AI coding-agent users working in Claude Code, Cursor, Codex, VS Code, Windsurf, or Claude Desktop
- Companies needing a single API across multiple hosted browser and extraction providers
- Teams automating research, extraction, and authenticated web workflows
- +1 more
AI Tools by Banana Peel
(1)
Banana Peel
Browser Agent Routing API
Discussions
No discussions yet
Be the first to start a discussion about Banana Peel
Products & Services
An OpenAI-compatible Responses-style API at /api/v1/responses. A caller supplies a task and optional URL; Banana Peel returns the result plus runner, timing, cost, fallback, and routing metadata.
Production routing that scores candidate runners using task/site success likelihood, cost, speed, and block risk, then can race or fall back to additional runners.
The official stdio package @banana-peel/mcp, exposing native tools for creating and polling browser tasks, submitting MFA input, listing runs, and viewing live runner rankings.
The @banana-peel/cli package can create an agent account, store credentials locally, provide a claim URL, and configure supported agent environments such as Claude Code, Cursor, Codex, VS Code, Windsurf, and Claude Desktop.
Market Position
Banana Peel positions itself as an OpenRouter-like aggregation and routing layer for browser agents rather than as a single browser infrastructure provider. Its differentiation is one API/key/bill across a broad runner catalog, learned per-domain routing, and parallel fallback; named alternatives in its own documentation and integrations include Browserbase, Steel, Skyvern, Hyperbrowser, Browser Use, and Playwright.
Founding Story
Banana Peel's initial product vision is to provide an 'OpenRouter for browser agents': one API, one key, and one bill instead of separate provider accounts, while using live per-domain success data to choose the execution runner most likely to complete a task.
Business Model
Revenue Model
Pay per executed runner attempt. Customers top up credits at face value; after welcome credits are used, each run is charged the selected runner's listed price plus a 5% Banana Peel platform fee. There are no seats, subscriptions, minimums, or platform fee during the initial welcome-credit balance.
Pricing Tiers
$10 welcome credits, 0% platform fee, all 250+ runner combinations, 7-day copied-data retention, and no card required.
All 250+ runner combinations; card payments via Stripe; 7-day copied-data retention; output normalization is an optional additional $0.10 per run.
Volume and contract terms, invoicing, negotiated retention, SSO, SLAs, regional execution controls, custom proxies, BYOK, and dedicated support.
Target Markets
- Developers and engineering teams building browser-agent applications
- AI coding-agent users working in Claude Code, Cursor, Codex, VS Code, Windsurf, or Claude Desktop
- Companies needing a single API across multiple hosted browser and extraction providers
- Teams automating research, extraction, and authenticated web workflows
- Enterprise customers requiring retention controls, SSO, SLAs, regional execution controls, or BYOK
- Browser-based web research and extraction
- Automating multi-step browser workflows
- Authenticated website tasks using stored credentials
- Tasks on sites with bot defenses or varying runner compatibility
- Integrating browser automation into applications through an OpenAI-compatible API
- Using browser-agent capabilities from coding-agent and MCP clients