What datawebot does
datawebot connects an MCP client to online shopping sources and returns product and seller information for agent workflows. Product searches can include titles, current and original prices, shop names, ratings, review counts, images, condition, links, and retrieval timestamps. Searches support country selection and can optionally be restricted to shop domains.
For a product returned by search, the offer lookup reports individual sellers, prices, totals including shipping, delivery information, seller ratings, and direct product-page links. This makes the server suitable for comparing where a product is sold rather than relying on a single retailer result.
The server also provides Extraction Jobs for search results and product pages. Supported coverage includes Amazon sites in 10 countries, Walmart US, reviewed stores such as Allbirds, Ring, Arlo, ecobee, and iFixit, plus more than 11,000 store domains listed in the endpoint directory. Availability and qualification should be checked for each job contract, especially for stores added through automated probing.
How it works
The hosted endpoint is available at https://mcp.datawebot.com/mcp over streamable HTTP. A client connects to that URL and authenticates before calling tools. OAuth 2.1 uses the MCP authorization flow with PKCE, dynamic client registration, refresh tokens, and email approval. Clients and scripts that cannot complete the email flow can send an API key as Authorization: Bearer <your API key>.
Call discover to see the tools available to the session, including tools added after the session began. Search with search_products, then pass a selected product to get_product_offers when seller-level comparison is needed. Extraction workflows use list_job_blueprints, submit_job, get_job, and get_results.
Responses are best-effort snapshots captured at retrieval time. A response also reports the remaining free daily allowance for the account. New accounts receive 200 trial credits, and additional credits are available through the dashboard.
Setup and configuration
There is no local server package to install because datawebot is hosted. Configure an MCP client with the remote URL, then approve the sign-in email. Cursor, VS Code, Cline, LM Studio, Claude, Gemini CLI, Claude Code, and other clients can use the endpoint through their MCP settings or integrations.
Claude Code can install the datawebot plugin, which also includes the server and a datawebot agent skill. Gemini CLI can install the repository as an extension. For scripts, n8n, and agent frameworks, create an API key in the datawebot dashboard and store it in an environment variable or credential store rather than embedding it in source code.
A direct MCP configuration uses the streamable HTTP URL. Cline configurations should set the transport type to streamableHttp; otherwise it may use legacy SSE. LM Studio requires version 0.3.17 or later for the documented setup.
Tools and capabilities
search_products: finds products across online shops and returns structured product metadata.get_product_offers: compares sellers for a product found through search.discover: lists tools currently available to the session.list_job_blueprints: shows available extraction-job contracts and their coverage.submit_job: starts an extraction job for supported search results or product pages.get_job: checks an extraction job's status.get_results: retrieves completed extraction results.
Extraction coverage can change, and stores marked as probe-qualified may still be awaiting live production acceptance. Check each blueprint's qualification and availability fields before building a workflow around it.
Limitations and notes
Prices are not verified or guaranteed to remain current. The server does not maintain price history, schedule recurring checks, or send alerts. An agent or external script must save results when historical tracking is required. Store coverage and job availability should be evaluated at runtime through the available blueprints.
A free daily allowance applies per account. Extraction Jobs consume credits, with additional credit packs available from the dashboard. OAuth requires a client and workspace that support custom MCP authentication; API keys provide an alternative for compatible scripts and frameworks.
