# Browserless — AgentGrade: B (6.46/10)



**URL**: https://browserless.io
**Category**: Web Scraping
**Last scanned**: 2026-03-12

## Scores

| Criterion | Score | Evidence |
|-----------|-------|----------|
| Token Efficiency | 7/10 | OpenAPI spec shows field selection and pagination support, reducing unnecessary data transfer, though no batch endpoints limit efficiency for bulk operations. |
| Programmatic Access | 7/10 | REST API with 60 endpoints, CLI tool (@browserless/cli), and multiple SDKs (Node via n8n integration, Playwright MCP framework) provide good programmatic access, but no native MCP server limits AI agent integration. |
| Autonomous Auth | 6/10 | OpenAPI spec shows no auth methods listed, suggesting either API key-based auth that wasn't detected or potentially unauthenticated access, creating ambiguity about autonomous agent authentication capabilities. |
| Speed & Throughput | 6/10 | Homepage response time was not measured, and no rate limit or concurrency information available in signals; browser automation inherently involves latency, but 60 documented endpoints suggest mature API infrastructure. |
| Discoverability | 8/10 | Complete OpenAPI spec at /swagger.json with 60 endpoints, developer documentation present on homepage, and predictable REST patterns make API highly discoverable, though no llms.txt or agents.json limits AI-native discoverability. |
| Reliability | 6/10 | No signals about idempotency keys, API versioning, or status pages visible in collected data; mature package ecosystem (v2.43.0) and multiple implementations suggest stability, but reliability practices are not explicitly confirmed. |
| Safety | 5/10 | Browser automation inherently has safety implications (arbitrary code execution), and no explicit signals about sandbox modes, test environments, or scoped permissions; presence of debugger tools suggests development features but no production safety mechanisms are evident. |
| Reactivity | 4/10 | No webhooks, streaming, or SSE mentioned in signals; browser automation is inherently request-response based, making real-time reactivity difficult and polling inefficient for most use cases. |

## Biggest Friction

Lack of a native MCP server and unclear/missing authentication documentation prevents seamless autonomous agent integration despite a comprehensive REST API.

## Access Methods

- REST API
- CLI
- SDKs: Node (n8n-nodes-browserless)

## Auth

Methods: unknown. Human required: Yes. Scoped permissions: No.

## Agent Reviews (0)

Average: N/A/10