Resources · Claude connector

Claude LatchBio connectorWhat Claude can do in your LatchBio account.

The Claude LatchBio connector exposes 9 tools. 8 of them read your Latch workspace (files, workflows, executions, logs) and 1 starts a workflow run. Below: what Claude can see, the one action it asks about first, and what a launch costs you.

Verified Trustpilot reviews · AI, automation & growth agency

Overview

What changes when Claude can reach your Latch workspace

With the connector, you ask in plain words and Claude works in your Latch workspace: it lists folders, finds a pipeline, reads its parameters, starts a run once you agree, and reports back on status and logs, all inside the conversation. Latch presents it as a way to run serverless pipelines written in Nextflow or Python without leaving Claude.

Know what is in a folder before you touch it. Claude calls list_files to show what sits in a directory of Latch Data, then get_file to hand you a Console link or a temporary download link for the file you pick.

Go from "which pipeline?" to a run. list_workflows surfaces the workflows of your workspace plus public ones, get_workflow_schema shows which parameters a launch needs, and launch_workflow starts it with values that match that schema. Latch says Claude asks your permission before that launch by default.

Follow a run without opening the Console. list_executions and get_execution give status, task nodes and result files; get_task_logs brings the logs of a failing task into the chat.

What Claude will not do here: read the contents of a file through list_files (Latch says that tool returns names, not contents), edit or delete your data, or start anything on its own schedule. Every run you launch through Claude is billed like a run started from the Console. The connector also does not replace your Latch login: its sign-in only serves Claude, not the SDK. For other connectors documented the same way, the Integrations hub lists them.

Vocabulary

Five words to know before connecting

Claude's vocabulary for this connector, in one minute.

Connector
A link you set up once between Claude and an account you already have, here Latch, so Claude can work in it while it answers you.
Tool
One named action the connector gives Claude, such as listing executions. Claude picks the ones it needs on its own while replying.
Authorization
Latch's own sign-in screen, where you grant the access Claude will use. It is given once per person and can be withdrawn later.
Approval
The confirmation Claude waits for before doing something that changes your account, such as starting a workflow run on Latch.
MCP
The open standard connectors are built on: it is what lets an assistant like Claude talk to an outside service such as Latch.
Connect

Connect LatchBio to Claude in three steps

  1. 01

    Open the connector list

    In Claude's settings, go to Customize, then Connectors, and find LatchBio. On a Team or Enterprise workspace, an Owner or Primary Owner switches the connector on first; after that, each member connects on their own.

  2. 02

    Sign in to Latch

    Click Connect on the LatchBio row. Latch opens its own window, where you log in with your Latch account. If the link breaks later, use Disconnect and connect again.

  3. 03

    Approve the access screen

    Read the authorization screen before you accept. It is Latch's screen, not Claude's, and it sets what the access covers. You can also withdraw that access later from the Latch side.

Tools

The 9 LatchBio tools, read versus write

LatchBio gives Claude 9 tools: 8 that read your account, 1 that change something in it.

Two groups: what Claude reads in your workspace, and the single tool that starts something. Names stay exactly as Claude displays them.

  • 8 read
  • 1 write

What Claude reads (8)

8 tools

Eight tools that look at data, workflows and runs without changing them.

list_files

Shows the immediate contents of one directory in Latch Data, so you see the names of what sits there. It stops at that level and never returns what is inside the files.

When it helps
you are not sure which sequencing batch landed in a shared folder this morning.

Sourcewiki.latch.bio · October 1, 2026 ↗

get_file

Hands back a way to reach one stored file: either a link that opens it in the Latch Console or a presigned URL to download it directly.

When it helps
a collaborator needs a quick link to the output table of last week's analysis.

Sourcewiki.latch.bio · October 1, 2026 ↗

list_workflows

Finds the workflows you could launch: the ones built in your own workspace and the public ones Latch makes available to everyone.

When it helps
you remember there is an alignment pipeline somewhere in the lab's workspace but not its exact name.

Sourcewiki.latch.bio · October 1, 2026 ↗

get_workflow_schema

Fetches the launch details of one workflow along with its parameter schema, which tells Claude what values a run expects before anything starts.

When it helps
you want to check the required inputs of a pipeline before preparing your samples.

Sourcewiki.latch.bio · October 1, 2026 ↗

list_executions

Returns the workflow runs of a workspace, optionally narrowed by name, workflow or status, so Claude can spot failed or running jobs in one pass.

When it helps
Monday morning, you want to know which of the weekend's runs failed.

Sourcewiki.latch.bio · October 1, 2026 ↗

get_execution

Pulls the detail of one run: its status, its task nodes, the result files and a Console URL to open it in the browser.

When it helps
a run is marked as done and you need the path to its outputs.

Sourcewiki.latch.bio · October 1, 2026 ↗

get_task_logs

Brings in the logs of a task, either inline in the conversation or as a presigned URL when you want the complete file.

When it helps
one step of a run crashed and you want Claude to read the error with you.

Sourcewiki.latch.bio · October 1, 2026 ↗

list_workspaces

Lists every workspace your account can reach, the default one included. It only reports which workspaces exist for you and leaves them as they are.

When it helps
you belong to several labs on Latch and want to be sure the files and runs you are about to ask for come from the right one.

Sourcewiki.latch.bio · October 1, 2026 ↗

What Claude starts (1)

1 tool

One tool that launches a billed workflow run, with your permission asked first by default.

launch_workflow

Asks you first

Starts a run of a workflow with parameter values that match its schema. It is the only tool here that sets something in motion on Latch.

What Claude asks for
Latch states that, by default, Claude requests your permission before launching a workflow execution.
When it helps
the samples are uploaded and you want the pipeline running before you leave.
Watch out
the run is charged exactly as if you had launched it from the Console or SDK.

Sourcewiki.latch.bio · October 1, 2026 ↗

Approvals

What Claude asks before acting

By default, Claude pauses and asks for confirmation before each action it carries out on an account for someone.

For LatchBio, the publisher spells it out for the one tool that matters: a workflow launch waits for your yes. On Team and Enterprise workspaces, owners decide whether a member may let some actions pass without a prompt each time, and they can limit what a connector does for the whole organization. Claude also uses your rights and nothing more: a workspace you cannot open on Latch stays closed to it.

Plans

Which plans can use it

No official source publishes plan availability connector by connector, and none of the 819 sheets in the directory shows it.

The published rule is general: remote connectors are open to all users on Claude, Cowork, Claude Desktop and mobile. On Team and Enterprise, an Owner or Primary Owner opens the connector for the organization before members connect. For LatchBio's live state, check its sheet in the official directory. Separately, you need a Latch account.

Limits

Where the LatchBio connector stops

A connector is not an automation. Claude calls these tools while it answers you: no run starts by itself when new files land in Latch Data.

Everything known about this connector comes from its publisher: no Claude help page covers it. The partner badge is not a security audit, and Anthropic states it neither picks the tools a publisher exposes nor guarantees how they behave. The tool list is also a floor rather than a ceiling, since an administrator can open actions no public sheet shows. Anthropic's own advice on every sheet applies here too: only connect a publisher you know.

Need help

Need help connecting LatchBio to Claude?

A person reads every message.

FAQ

Questions about the Claude LatchBio connector

01What can Claude do with the LatchBio connector?
Claude can browse your Latch workspace and run pipelines from the conversation. Its nine tools list your workspaces, browse directories of Latch Data, return a Console link or download URL for a file, find workflows and their parameter schemas, launch a run, and follow executions down to the task logs. Everything happens while Claude answers you, using the access you granted on Latch's sign-in screen. Nothing runs on a schedule or reacts to new data on its own.
02Can Claude change or delete anything in Latch?
Only one tool acts on your account: launch_workflow, which starts a workflow run with parameters that match its schema. The eight other tools read files, workflows, executions and logs. No published tool edits files, deletes data or cancels a run. That list is what the directory sheet shows today, and an administrator can open actions that appear on no public sheet, so treat it as the current floor rather than a permanent guarantee.
03Does Claude ask before launching a workflow?
Yes. Latch's documentation states that, by default, Claude asks for your permission before it launches a workflow execution. This matches Claude's general rule of confirming before acting on an account for someone. On Team and Enterprise workspaces, owners decide whether members may let some actions run without a prompt each time. Keep the prompt in mind, because every launch is billed the same way as a run started from the Latch Console or SDK.
04Which Claude plans include the LatchBio connector?
No official source says which plans include it, connector by connector, and none of the 819 directory sheets lists plan availability. The general rule is that remote connectors are open to all users on Claude, Cowork, Claude Desktop and mobile, and that on Team and Enterprise an Owner or Primary Owner enables a connector before members can connect. The LatchBio sheet in the official directory is the only place that shows its current state.
05Can Claude see every workspace on Latch?
No. Claude only reaches what your own Latch account reaches, and the list_workspaces tool returns the workspaces you can access, the default one included. A colleague with access to more labs would see more through the same connector. On Team and Enterprise workspaces, an owner can also narrow what the connector may do for the whole organization, for example keep reading and block launches, and members cannot undo that setting.
06Does launching a workflow from Claude cost extra?
No separate price exists for the connector, but runs are not free either. Latch says any action performed through it is charged normally: a workflow launched from Claude costs the same as one launched from the Latch Console or SDK. That is why the permission prompt before launch_workflow matters: it is your last look at a run before it is billed. Reading tools such as list_executions or get_task_logs do not start anything. Latch also notes that you need a Latch account to use the connector at all.
07Claude or an automation tool for Latch pipelines?
They solve different problems. The connector works inside a conversation: you ask Claude which runs failed, read the logs together and decide to launch a new run, with your permission requested first. It never starts anything by itself when files arrive. If you want a run to start automatically every time new sequencing data lands, that is a job for a scheduled automation outside this connector, not something Claude does on its own.