Benchling MCP Server

Official remote MCP server for Benchling's R&D cloud. Lets AI agents query scientific data including notebook entries, experiments, sequences, inventory, and workflows.

Healthcare & Life Sciences by Benchling OAuth2 active

Connect (Remote Server) Docs

Overview

Benchling MCP is the official, remote Model Context Protocol server hosted by Benchling as part of its AI Connectors offering. It lets external AI tools such as Claude, ChatGPT, or custom agents securely query a tenant's Benchling R&D data and return structured, grounded answers. Scientists can stay in their preferred AI environment while pulling experimental records, sequences, inventory, schemas, and workflow tasks directly from Benchling.

The server is tenant-scoped: each customer gets a URL of the form https://<tenant>.mcp.benchling.com/mcp and connects via OAuth 2.1 with delegated authentication. API key auth is not supported on the remote MCP server. Access enforces the same permissions, governance, and audit controls as the Benchling web app, so AI agents only see what the authenticated user is allowed to see.

Benchling MCP Server entered general availability after a customer beta, and the broader AI Connectors platform (MCP Client and MCP Directory, with pre-built connectors for Elicit, GXL, Quilt, and Seqera, plus Notion, SharePoint, and Snowflake) is rolling out in May 2026. Tenants must have Benchling AI (Deep Research) and the V3 APIs enabled.

Tools

Tool	Description
`query_benchling`	Run a natural-language query across the tenant's Benchling data and return structured results grounded in notebook entries, experiments, registry entities, and inventory.
`search_experiments`	Locate and filter experimental records, notebook entries, and studies in Benchling.
`get_entity`	Retrieve detailed information for a specific Benchling entity such as a DNA or protein sequence, plasmid, sample, or result row.
`retrieve_data`	Fetch structured data from notebooks, results tables, schemas, and inventory items.

Setup Guide

Benchling hosts the MCP server. There is nothing to install locally. You only need to point your MCP client at your tenant URL and authenticate with OAuth.

Prerequisites

A Benchling tenant (paid/enterprise plan)
Benchling MCP, Benchling AI (Deep Research), and V3 APIs enabled on the tenant by an admin
An MCP client that supports remote servers with OAuth 2.1 (e.g. Claude Desktop, Cursor, ChatGPT)

Server URL

https://<tenant>.mcp.benchling.com/mcp

Replace <tenant> with your Benchling subdomain (the same one you use at https://<tenant>.benchling.com).

Claude Desktop / Cursor config

{
  "mcpServers": {
    "benchling": {
      "url": "https://<tenant>.mcp.benchling.com/mcp"
    }
  }
}

Authentication

Use OAuth 2.1 with delegated authentication. If your client only exposes "OAuth 2.0," select that.
Prefer Dynamic Client Registration (DCR) or Dynamic Discovery when available.
Redirect / callback URI format: https://<tenant>.benchling.com/1/api/llm/mcp/oauth/callback.
API key authentication is not supported by the remote MCP server.

On first connection your client will open a browser to your Benchling tenant where you log in and approve the MCP scopes. Tokens are then managed by the client.

Use Cases

Ask Claude or ChatGPT to summarize the latest results from a specific antibody or cell line project pulled live from your Benchling notebook entries.
Generate a draft methods or results section for a presentation by querying experiment records and registry entities without leaving the AI chat.
Look up DNA / protein sequence details, plasmid maps, or sample metadata by registry ID inside an agentic workflow.
Search across notebook entries and inventory to answer questions like "which lots of antibody X were used in assays last quarter?"
Combine Benchling R&D data with external sources (literature via Elicit/GXL, S3 data via Quilt, pipelines via Seqera) in a single AI session.

Example Prompts

"Find all notebook entries from the CRISPR screening project tagged 'hit confirmation' in the last 30 days and summarize the key results."
"Get the sequence and annotations for plasmid pBEN-1042 from Benchling and check whether it contains a Kozak sequence."
"List the antibody clones in our registry with binding affinity below 5 nM and link the supporting experiments."
"Pull the inventory locations and remaining volumes for all samples in workflow task WT-2381."
"Draft a Q2 progress update for the oncology program using data from Benchling experiments in the Oncology project."

Pros

Official, vendor-hosted server with no infrastructure to run or maintain.
Tenant-scoped OAuth 2.1 authentication respects existing Benchling permissions, governance, and audit logging.
Provides AI-grounded access to the canonical R&D system of record (notebooks, registry, inventory, workflows) instead of disconnected exports.
Designed to work alongside Benchling's MCP Client and Directory for combining R&D data with literature, S3 datasets, and bioinformatics pipelines.

Limitations

Requires an enterprise Benchling tenant with Benchling AI (Deep Research) and V3 APIs enabled; not available on free or trial accounts.
Public, detailed tool reference is limited; exact tool names and parameters are not fully documented outside the Benchling help center (which is gated).
No API key auth; clients must support OAuth 2.1 / 2.0 with delegated flows, which some MCP clients still handle inconsistently.

Alternatives

longevity-genie/benchling-mcp (community, open-source Python MCP server using the Benchling SDK and API keys; useful if you cannot use the hosted OAuth server).
Benchling SDK / V2 REST API called directly from a custom MCP server or agent if you need finer-grained control than the hosted MCP exposes.
Generic FHIR or LabKey MCP servers for adjacent lab data systems, though they do not cover Benchling's registry and notebook model.