Bright MCP
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Bright MCPsearch the web for electric car reviews"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Agent-oriented Bright Data capabilities over MCP, built on Bun. The seven-tool all profile separates search, ranked source discovery, exact reading, extraction, research, maintained dataset discovery, and execution. It pages complete pages and upstream snapshots as resources and renders structured results in a transient React MCP workbench.
The full seven-tool contract remains at /mcp. Entitlement-aligned installs can
use stable three- or two-tool surfaces at /mcp/web, /mcp/deep-lookup, or
/mcp/marketplace; Scraping Browser is a separate four-tool surface at
/mcp/browser. Tool lists never change after initialization based on a probe.
Install
Set BRIGHTDATA_API_KEY in the client environment first.
Plugin
Codex:
codex plugin marketplace add dunkeln/bright_mcp --ref main
codex plugin add bright@brightClaude Code:
claude plugin marketplace add dunkeln/bright_mcp
claude plugin install bright@brightMCP
Codex:
codex mcp add bright --url https://bright-mcp.onrender.com/mcp \
--bearer-token-env-var BRIGHTDATA_API_KEYClaude Code:
claude mcp add --transport http bright https://bright-mcp.onrender.com/mcp \
--header "Authorization: Bearer ${BRIGHTDATA_API_KEY}"Cursor (~/.cursor/mcp.json):
{
"mcpServers": {
"bright": {
"url": "https://bright-mcp.onrender.com/mcp",
"headers": { "Authorization": "Bearer ${env:BRIGHTDATA_API_KEY}" }
}
}
}The key is forwarded over HTTPS and kept only in a bounded in-memory cache; Bright MCP does not persist it. Available live capabilities follow the products enabled on that Bright Data account.
Choose the narrowest surface your account and workflow need:
Endpoint | Tools | Bright Data access | Authorization |
| All seven data tools | SERP, Discover, Web Unlocker, Deep Lookup, Marketplace as used | Bearer API key |
|
| SERP + Discover + Web Unlocker | Bearer API key |
|
| General Deep Lookup | Bearer API key |
|
| Account-visible Marketplace datasets | Bearer API key |
| Four | Scraping Browser | HTTP Basic with Browser API username/password |
Choose among the seven data tools by intent:
Sources | Needed result | Tool |
Unknown | Compact links and summaries |
|
Unknown, goal-constrained | Ranked source shortlist |
|
Known URLs | Readable page evidence |
|
Known URLs | Exact source HTML |
|
Known URLs | Temporary named fields |
|
Unknown | Sourced structured records |
|
Maintained vertical data | Typed records |
|
See SETUP.md for local development, credentials, live checks, and hosted authorization.
Related MCP server: WebQuest MCP
Benchmarks

Both MCPs completed 29 of 30 workflows; the meaningful separation begins after completion, in the quality and efficiency of the answer.

Bright leads every blind-scored quality dimension, with its clearest gains in fulfillment, grounding, source quality, and actionability.

The blind judge preferred Bright 17 times versus 3 for BrightData, while preserving 10 genuine ties.

Structured Marketplace work is where Bright's richer answers most clearly repay their token budget; simpler web tasks remain the efficiency target.

Bright converts tokens into successful Marketplace execution efficiently, while the web workflows show where tighter routing can recover more value.

Bright batches known-page reading into one call and makes Marketplace discovery explicit; historical search follow-ups expose the routing behavior now being tightened.
At the same 97% completion rate, Bright MCP delivered 4.51/5 judged quality versus 3.78/5, used slightly fewer tokens, and won the blind preference 17–3 with 10 ties.
Method and tables · current-entitlements Acquire + Operate profile · openrouter/anthropic/claude-haiku-4.5 · 10 runs/case · 2026-07-22.
WIP capabilities
Recurring delivery is intentionally excluded from the current benchmark score. Bright MCP can discover and run datasets, but it cannot yet create a durable refresh schedule; the case returns when it can execute delivery instead of only describing that boundary.
A/B
Dimension | BrightData MCP | Bright MCP |
Model-visible tools | 60+ maximum | 7 data / 4 browser |
Browser tools | 14 | 4 |
Dataset tools | One per dataset | Discovery + execution |
Dataset catalog | Tool inventory grows with products | Caller-scoped catalog behind discovery |
Research | Search then agent-managed scraping | Dedicated sourced research with cost gates |
Large results | Returned through tool calls | Lazy, principal-bound snapshot resources |
Runtime/toolchain | Node + npm + Vite | Bun-native |
Production dependencies | 7 plus UI dependencies | Roughly 6–8, profile-dependent |
API-specific code | Repeated across tools | Central adapters |
Polling implementations | Repeated | One shared mechanism |
Schema definitions | Repeated per tool | Catalog/operation-driven |
Credentials | API key in endpoint configuration | BYOK from the client environment; never server-funded |
Resources/tasks | Limited | First-class |
Security/session controls | Relatively implicit | Explicit and bounded |
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Latest Blog Posts
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/dunkeln/bright_mcp'
If you have feedback or need assistance with the MCP directory API, please join our Discord server