Reported issues for brightdata-mcp

Pod holds 23 of 45 problems reported by people outside the maintainer team. Issues filed by the project's own owners, members and collaborators are excluded entirely — a maintainer's release checklist is not a warning to a prospective user.

Back to brightdata-mcp.

Most discussed

Remote MCP (mcp.brightdata.com) returns empty results while showing "success" in event logs

Similar to #69, but affecting the remote MCP (mcp.brightdata.com/mcp?token=...) rather than the local npx version.

Symptoms

• MCP event logs show status: success for all scrape_as_markdown requests
• Tokens Used: 0 for all requests
• URL column is empty in event logs
• Requests never return content to the client (Claude Code in my case)
• Client times out waiting for response

Event Log Evidence

Date,Tool,Event,Client Name,URL,Status,Tokens Used
2026-02-02T11:10:14.

[Read the thread](https://github.com/brightdata/brightdata-mcp/issues/106) · 2026-02-02 · closed · external user · 7 comments

### I can't connect my the brightdata MCP to ChatGPT

I have tried to connect your MCP service to chatGPT on at least 10 occasions but I only get error messages in response (see attached images). 

<img width="1470" height="956" alt="Image" src="https://github.com/user-attachments/assets/15ba6f5c-c8f5-4400-aa28-83f2f52f2445" />

<img width="1470" height="956" alt="Image" src="https://github.com/user-attachments/assets/b048719b-b3c4-4320-bf45-b56588d2480d" />

My account is listed under the following email: [mark.lewyn@gmail.com](mailto:mark.lewyn@g

[Read the thread](https://github.com/brightdata/brightdata-mcp/issues/72) · 2025-10-01 · closed · external user · 7 comments

### control over what search_engine returns?

the Bright Data api itself (which the mcp wraps) supports very granular (object-based) control over search_engine results. 
I understand that the go-to way to expose stuff to LLMs is markdown so you chose to expose the markdown as is. 
the thing is that the search_engine markdown is full of "trash": ads, "feedback", AI suggestions...
I don't want my agent to be exposed to all of that. I want it to get the results, and that's it.
any chance you could do a "sterile" search_engine, that returns onl

[Read the thread](https://github.com/brightdata/brightdata-mcp/issues/66) · 2025-09-04 · closed · external user · 5 comments

### Getting Snapshot is building, try again in 30s

I am getting **snapshot is building** in the response after few attempts and there is no way for me to wait till the snapshot is fully build. Could you please help me in handling this?

[web_data_crunchbase_company] executing {"url":"https://www.crunchbase.com/organization/zekith-india-private-limited"} [web_data_crunchbase_company] triggered collection with snapshot ID: s_mdzulo7o2opv5236pw [web_data_crunchbase_company] snapshot not ready, polling again (attempt 1/600) [web_data_crunchbase_

Read the thread · 2025-08-06 · closed · outside contributor · 9 comments

Amazon Product Search Failing - No Results Returned

Description: When attempting to use the web_data_amazon_product_search function, the tool returns "No result received from client-side tool execution" error. Steps to Reproduce:

Call the web_data_amazon_product_search function with the following parameters:

keyword: "glass cleaner" pages_to_search: "1" url: "https://www.amazon.com/"

The function returns an error instead of search results

Expected Behavior: The function should return a list of glass cleaner products from Amazon search result

Read the thread · 2025-07-29 · closed · external user · 28 comments

Add field filtering parameters to Bright Data MCP functions to reduce token consumption

Issue Description:

Problem Summary

The current Bright Data MCP implementation returns complete product data (8-10k tokens per product) with no ability to filter specific fields. This causes token limit issues when processing multiple products, making the tool impractical for analyzing 10+ products in a single conversation.

Current Behavior

All Bright Data MCP functions return complete, unfiltered data:

# Current function signatures (no filtering options):
web_data_amazon_

[Read the thread](https://github.com/brightdata/brightdata-mcp/issues/40) · 2025-07-26 · closed · external user · 18 comments

### Docker

This looks great would it be possible to get this as a Docker image?

[Read the thread](https://github.com/brightdata/brightdata-mcp/issues/25) · 2025-06-16 · closed · external user · 6 comments

### Browser API tools not working

It seems that the Browser API tools do not work.

Even on the playground (https://smithery.ai/server/%40luminati-io%2Fbrightdata-mcp/tools) some of the tools (scraping browser get text or scraping browser link) do not load although the API token and Browser Auth was configured. From testing it seems only a few tools do not work and others such as get screenshot do work. It would be really useful to be able to use the tools as for example giving an AI Agent the whole HTML using web unlocker tools

[Read the thread](https://github.com/brightdata/brightdata-mcp/issues/9) · 2025-05-11 · closed · external user · 8 comments

## Most recent

### search_engine_batch with Google engine returns raw HTML instead of JSON intermittently

## Description

When using the MCP tool  with , the response intermittently returns raw HTML instead of parsed JSON. This happens roughly 50% of the time.

## Error

Unexpected non-JSON response from Bright Data for search_engine_batch

The error originates inside the compiled .mcpb bundle (), not in any downstream consumer. The bundle appears to pass through the raw Google SERP HTML when the underlying API returns a non-JSON response, without validating or retrying.

## Related

- Exist

[Read the thread](https://github.com/brightdata/brightdata-mcp/issues/167) · 2026-08-24 · open · external user · 0 comments

### Security scan results for brightdata-mcp — MCPSafe AIVSS 81/100 (Grade B)

Hi team 👋

I ran a free deep security scan of **brightdata/brightdata-mcp** using [MCPSafe](https://mcpsafe.io) — a purpose-built scanner for MCP servers using a 5-LLM consensus panel to detect prompt injection risks, over-scoped tool schemas, supply chain issues, and more.

## Results: 81/100 · Grade B

| Severity | Count |
|----------|-------|
| 🔴 Critical | 0 |
| 🟠 High | 0 |
| 🟡 Medium | 15 |
| 🟢 Low | 0 |

**Summary:** 15 medium-severity findings around web data collection tool schemas and 

[Read the thread](https://github.com/brightdata/brightdata-mcp/issues/137) · 2026-05-12 · open · external user · 0 comments

### Proposal: add a Perplexity Search MCP tool as an alternative search source

Hi Bright Data team — thanks for maintaining this MCP server! I'd like to propose adding a `perplexity_search` MCP tool that wraps Perplexity's Search API (`POST https://api.perplexity.ai/search`) as an alternative/complementary search source alongside `search_engine`.

### Why

The current `search_engine` tool routes through Bright Data's SERP infrastructure, which is great for grounded SERP scraping. Perplexity's Search API is a different shape — it's a curated, ranked web index optimized for 

[Read the thread](https://github.com/brightdata/brightdata-mcp/issues/136) · 2026-04-29 · open · external user · 1 comment

### Add policy enforcement to prevent runaway scraping costs

This server exposes 60+ tools including browser automation and batch scraping on a pay-per-use billing model:

- `scrape_batch` -- bulk scraping across many URLs (high cost per operation)
- `scraping_browser_click_ref` -- browser automation on live sites
- Browser session tools -- interact with authenticated sites
- Web data API tools -- structured data extraction at scale

An agent in a loop calling batch scraping tools can generate significant unexpected costs. Browser automation tools can int

[Read the thread](https://github.com/brightdata/brightdata-mcp/issues/132) · 2026-04-12 · open · external user · 0 comments

### Add country flag to scrape_as_markdown & scrape_as_html

Allow country flag for problematic sites that require a specific geo location.

[Read the thread](https://github.com/brightdata/brightdata-mcp/issues/131) · 2026-04-06 · closed · outside contributor · 0 comments

### [RFC] Agent-native payment gating for Bright Data MCP — pay-per-scrape via x402/MPP

## What this proposes

Adding pay-per-call payment gating to Bright Data MCP so AI agents can pay for web data access directly — without requiring a human-managed Bright Data account for every deployment.

Bright Data already has per-request pricing. This extends that model to agent-native payments: agents pay per scrape/search via x402 (USDC on Base) or MPP (Stripe's Machine Payments Protocol), settled directly to the operator — no credit card, no monthly billing cycle, no human in the billing 

[Read the thread](https://github.com/brightdata/brightdata-mcp/issues/126) · 2026-03-31 · open · external user · 1 comment

### Outdated @modelcontextprotocol/sdk (<=1.25.3) - 3 high severity vulnerabilities

Hi,

Running `npm audit` on `@brightdata/mcp@2.9.2` reports 3 high severity vulnerabilities from `@modelcontextprotocol/sdk@<=1.25.3`:

1. **ReDoS vulnerability** - [GHSA-8r9q-7v3j-jr4g](https://github.com/advisories/GHSA-8r9q-7v3j-jr4g)
2. **Cross-client data leak** via shared server/transport instance reuse - [GHSA-345p-7cg4-v4c7](https://github.com/advisories/GHSA-345p-7cg4-v4c7)
3. **DNS rebinding** not enabled by default - [GHSA-w48q-cv73-mx4w](https://github.com/advisories/GHSA-w48q-cv73-m

[Read the thread](https://github.com/brightdata/brightdata-mcp/issues/123) · 2026-03-28 · closed · external user · 1 comment

### Looks like some packages has expired

Here's the log of the error i'm currently having:

2026-03-19T17:21:32.400Z [Bright Data] [info] Initializing server... { metadata: undefined } 2026-03-19T17:21:32.419Z [Bright Data] [info] Using MCP server command: /Users/kap/.nvm/versions/node/v15.14.0/bin/npx with args and path: { metadata: { args: [ '@brightdata/mcp', [length]: 1 ], paths: [ '/Users/kap/.nvm/versions/node/v15.14.0/bin', '/Users/kap/.nvm/versions/node/v16.20.0/bin', '/Users/kap/.nvm/versions/no

Read the thread · 2026-03-19 · closed · external user · 1 comment

[Bug] Remote MCP connection fails with HTTP 400: zone "mcp_unlocker" not found when using Free tier with empty Zone configuration

I am trying to set up a remote MCP server using the BrightData endpoint via the Llama.cpp connector. However, the connection fails with an HTTP 400 error stating zone "mcp_unlocker" not found.

Configuration Details:

• Plan Type: I have only selected the Free options in the dashboard.
• Zone Configuration: The field labeled "If you want to use a specific zone, configure it here" was left completely empty (I did not enter any zone name).
• Account ID: `hl_8ae1a

Read the thread · 2026-02-28 · closed · external user · 3 comments

Markdown minification for scrape_batch

Hey,

Just wondering if the same nice tweak you made for scrape_as_markdown (using remark to strip e.g. base64 images) could be made for scrape_batch too in an effort to reduce token count when batch scrapes are performed?

Thanks!

Read the thread · 2026-02-27 · closed · external user · 2 comments

Security scan results for your MCP server

Hey! 👋

I ran a security scan on this MCP server using MCPSafe and found a few items that might be worth reviewing.

Why this matters: MCP servers run with host permissions and are trusted by AI agents. Security issues can be exploited through prompt injection.

What I found: Your server scored in the mid-range on our security analysis. The detailed report shows specific patterns that could be hardened.

Free offer: I'd be happy to give you a complimentary Team

Read the thread · 2026-02-03 · closed · external user · 2 comments

"Zone has reached usage limit" Error when using as an n8n Tool

Dear Brightdata team,

We use Brightdata MCP as a Tool in N8N.
My team and I experience the "Zone has reached usage limit" often. Here is the full response body: <!doctype html><h1>Webpage not available</h1><p>The webpage could not be loaded because:</p><p>Zone has reached usage limit</p> The emails that are associated with our Brightdata account are: m@portal.ai, h@portal.ai.

Here's what we've already tried:

1. we bought the $499/mo plan.
2. we enabled the pay-as-you option right after th

Read the thread · 2026-01-23 · closed · external user · 2 comments

Intermittent 502 Bad Gateway errors from MCP endpoint under moderate call volume (no retry guidance)

Hi Bright Data team,

We’re using the Bright Data MCP in an agent-based workflow (via n8n). Under moderate load (~100 MCP calls per workflow run), we occasionally receive:

502 Bad Gateway responses (HTML error page from gateway)

The error is rare per-call, but statistically guaranteed at scale. There is currently no documented retry / backoff guidance specific to MCP usage.

Questions:

1. Are intermittent 502s expected from the MCP gateway?
2. Do MCP endpoints support Retry-After headers?
3. I

Read the thread · 2026-01-19 · open · external user · 5 comments

Feature request: additional tools

Our Brightdata MCP-powered agent currently performs quite well, but often is unable to complete tasks, and I think it would help if it had some additional tools. In particular, I think the ability to execute arbitrary JavaScript (like playwright-mcp's browser_evaluate) could really help (the agent sometimes encounters shortcomings of ARIA snapshots, and is able to inspect the HTML instead but unable to interact with it). Something like browser_fill_

Read the thread · 2025-12-09 · closed · external user · 5 comments

Feature request: country-specific MCP server

I'd like to be able to use Brightdata's MCP server with country-specific proxies - specifically, UK-based proxies.

I know it's possible to use 'residential zone' country-specific proxies, and I'd like this feature to be available for the MCP server (hopefully it's not too much work to implement!).

Additional information on our use case:

• we are only using scraping_browser_* tools, and not, for example, web_data_* tools - I imagine it'd make less sense to have country-specific proxies for some

Read the thread · 2025-12-03 · closed · external user · 2 comments

The remaining reports are on the project's issue tracker.