Resources · Claude connector

Claude Kepler connectorWhat Claude can do in your Kepler account.

The Claude Kepler connector exposes 9 tools. 6 read your research runs, filings metadata and workbooks, 3 start, extend or stop a run in your Kepler account, and 0 are left undocumented. Below: how a run works, what Kepler needs first, and where it stops.

Verified Trustpilot reviews · AI, automation & growth agency

Overview

What changes once Claude can run Kepler research

Instead of pulling filings by hand and rebuilding tables, an analyst asks Claude a question about a listed company and Claude hands it to Kepler. Kepler works from SEC filings, earnings call transcripts and market data, then returns an answer where each figure carries its citation, plus workbooks when you asked for a model.

Get a cited answer on margins or growth. An equity researcher asks for two companies' gross margins over recent quarters: Claude calls run_financial_research, then waits on get_run_result, which returns the final result once the run ends.

Rework a model without starting over. With continue_research, Claude sends a follow-up into the same conversation, so Kepler keeps the full context of the earlier run while it adjusts the model.

Take the numbers further. get_workbook_data exports one sheet of a workbook the run produced as CSV, so Claude can chart or compare the cells in the chat.

What it does not do: Kepler states that, through the connector, access covers your own runs and results only; your account settings and other users' data stay out of reach. Runs are not instant, either: Kepler says one usually takes a few minutes. And nothing starts on its own, such as a run on each new filing; Claude acts only while it answers you. For scheduled jobs, the Integrations hub lists automation platforms.

Vocabulary

Five words before you connect

The vocabulary around this connector, in one minute.

Connector
A link you set up once between Claude and an account you already hold, so Claude can work in it while it answers you.
Tool
One named capability a connector gives Claude. Claude picks the ones it needs as it goes, and the directory sheet lists them by name.
Authorization
The sign-in screen of the service, where you grant the access Claude will use. It is given once per person and can be revoked.
Approval
The confirmation Claude waits for before an action that changes an account, shown in the conversation when it applies.
MCP
The shared standard behind connectors: it lets an assistant such as Claude work with an outside service like Kepler.
Connect

Connect Kepler to Claude in three steps

  1. 01

    Find Kepler in Claude

    Open Claude's settings, go to Customize then Connectors, and look Kepler up in the list. On a Team or Enterprise workspace, an Owner or Primary Owner enables it for the organization before members can each connect.

  2. 02

    Sign in to Kepler

    Click Connect on its row and log in with your Kepler account in the window Kepler opens. If the connection breaks later, Disconnect it and connect again from the same row.

  3. 03

    Read the authorization screen

    Check the screen before you accept. It is Kepler's, not Claude's, and it decides what the access covers. You can take that access back later from the Kepler side as well.

Tools

The 9 tools, split by what they do

Kepler gives Claude 9 tools: 6 that read your account, 3 that change something in it.

Six tools read, three act on your research runs. Names stay exactly as Claude shows them.

  • 6 read
  • 3 write

What Claude reads (6)

6 tools

Six tools that fetch results, check status or look up reference data, leaving your account as it is.

continue_waiting

Resumes the wait on a long run after the first wait came back unfinished, with the same blocking behaviour as the first wait.

When it helps
a deep three-statement model is still being built when the first wait ends, and you would rather not start it again.

Sourcedocs.kepler.ai · October 1, 2026 ↗

get_run_result

Waits for a run to finish while streaming its progress, then returns its final result.

When it helps
right after a model request, so the answer lands in the chat.
Watch out
a very long run can return still running; that is expected.

Sourcedocs.kepler.ai · October 1, 2026 ↗

get_workbook_data

Exports the full cells of one workbook sheet as CSV, for analysis beyond the tables shown in a result.

When it helps
a model came back with several tabs and you want Claude to chart the revenue sheet rather than read it from a table.

Sourcedocs.kepler.ai · October 1, 2026 ↗

is_run_done

Checks once, without waiting, whether a run has finished, and returns right away. It is a quick look, not a way to collect the answer.

When it helps
you started two runs and want to know which one is ready before reading anything.

Sourcedocs.kepler.ai · October 1, 2026 ↗

list_recent_runs

Lists your most recent Kepler conversations, 10 by default and at most 25, to find an earlier run to pick up again.

When it helps
you want yesterday's Netflix model back without searching the app or asking for the same model twice.

Sourcedocs.kepler.ai · October 1, 2026 ↗

lookup_company

Resolves a company and lists the filings and earnings call transcripts Kepler holds, instantly and without a run. It returns metadata only.

When it helps
you want to know whether the latest quarterly call is available before asking for analysis.
Watch out
for figures or quotes, a research run is needed.

Sourcedocs.kepler.ai · October 1, 2026 ↗

What Claude starts or stops (3)

3 tools

Three tools that start, extend or stop a run saved in your Kepler account. No source details their confirmation, so the general rule below applies.

cancel_run

Approval: see the rule

Stops a run that is still going, for instance after a mistyped ticker. Kepler says it is non-destructive: the conversation and partial results survive and can be resumed.

What Claude asks for
no source spells out a confirmation, so the general rule covers it.
When it helps
you launched a duplicate request by mistake.

Sourcedocs.kepler.ai · October 1, 2026 ↗

continue_research

Approval: see the rule

Adds a follow-up to an existing Kepler conversation, keeping the full context of the earlier run, to refine a model, question prior results or widen the research.

What Claude asks for
nothing specific is documented; rely on the approvals section further down.
When it helps
the model is right but you want an extra year of history added.

Sourcedocs.kepler.ai · October 1, 2026 ↗

run_financial_research

Approval: see the rule

Launches a new Kepler run from a question or a model request written in plain language, and returns immediately with a conversation identifier. The run is saved to your Kepler account.

What Claude asks for
no source describes a confirmation for this tool, so the general approval rule below applies.
When it helps
a portfolio manager wants a sourced comparison of two retailers before a meeting.

Sourcedocs.kepler.ai · October 1, 2026 ↗

Approvals

What Claude checks with you first

By default, Claude stops and asks before each action it takes on an account for someone. The prompt appears in the conversation when it matters.

On Team and Enterprise workspaces, owners decide whether a member may let some actions through without being asked every time. They can also limit what a connector may do for the whole organization, for example leaving reading on and switching the rest off. That setting binds everyone, and nobody lifts it from a personal account. Claude works with the rights of the person who connected; Kepler adds that the connector reaches only your own conversations and results.

Plans

Which plans include it

None of the 819 sheets in the official directory states its plan availability. That answer is published nowhere, connector by connector.

The general rule is what Anthropic publishes: remote connectors are open to all users on Claude, Cowork, Claude Desktop and mobile, and desktop extensions install on Claude Desktop. On Team and Enterprise, an Owner or Primary Owner opens a connector for the organization before members can each connect. The connector's sheet in the official directory is the only place that shows its current state.

Limits

Where this connector stops

A connector is not an automation. Claude calls these tools while it answers you, so no run starts when a company files a new report.

Only Kepler documents this connector; neither the Claude help center nor Anthropic's developer docs has a page about it. The tool list is an observed floor, not a promise, since an administrator can open actions that appear on no public sheet. The partner badge is not a security audit, and Anthropic says on every sheet that it neither chooses a publisher's tools nor guarantees their behaviour. Kepler also says your research conversations and their results are saved to your Kepler account.

Need help

Need help connecting Kepler to Claude?

A person reads every message.

FAQ

Questions about the Claude Kepler connector

01What can Claude do with the Kepler connector?
Claude can run financial research on public companies through Kepler and bring back cited answers. The directory sheet lists nine tools: one starts a run, one sends a follow-up, one cancels a run, three wait for or check a result, one lists recent conversations, one looks up a company's filings and transcripts, and one exports a workbook sheet as CSV. According to Kepler, every figure in an answer is traced to its source document.
02Can Claude change anything in my Kepler account?
Only your research runs. Three tools act on them: one starts a run, which Kepler saves to your account, one adds a follow-up to an existing conversation, and one stops a run that is still going without deleting it. Kepler states that access through the connector covers starting, continuing and cancelling your own runs and reading your own results, and that your account settings and other users' data stay out of reach.
03Does Claude ask before it starts a Kepler run?
No source describes a specific confirmation for the tools that start, extend or cancel a run, so the general rule applies: Claude asks before each action it takes on an account for someone. On Team and Enterprise workspaces, owners decide whether members may let certain actions run without a prompt each time, and they can restrict what a connector may do across the organization, a setting members cannot lift themselves.
04Which plans is it available on?
No official source publishes plan availability connector by connector, and none of the 819 directory sheets shows it. Anthropic's general rule is that remote connectors are open to all users on Claude, Cowork, Claude Desktop and mobile. On Team and Enterprise, an Owner or Primary Owner enables a connector for the organization before members connect. Kepler separately requires an account with MCP access enabled. For the live state, open the connector's sheet in the official directory.
05Does Claude see other people's Kepler research?
No. Kepler says the connector is scoped to you, as in the app: your assistant reaches only your own conversations and results, and it cannot change anyone else's data. In general, Claude works with the rights of the person who connected and nothing more. Kepler adds that the connector cannot change your account settings either, only start, follow and stop your runs. In addition, on Team and Enterprise an owner can narrow what the connector may do for everyone.
06Why does a Kepler answer take several minutes?
Because each run is asynchronous and reads primary sources before it answers. Kepler says a run usually finishes in a few minutes, while deeper models can take longer. Starting a run returns immediately, and the tool that fetches the result streams progress while it waits. Kepler notes that a very long run can come back still running, and the wait then resumes with a dedicated tool. A quick status check is also available when you only need to know whether it is done.
07Should I use Claude or an automation tool with Kepler?
They serve different needs, so it depends on the work. Claude is suited to analysis you steer: you ask a question, read the cited answer, then refine the model with follow-ups in the same conversation. An automation tool fits jobs that must run without you, such as a report after every earnings release, because a connector never starts on its own. Neither ranks above the other; they do different jobs.