OrgSDK

Product / Lake

A data workspace for your team's agents

From repository data to shared context.

Bring GitHub records into one workspace. Let authorised agents explore the evidence, so your team can spend its time on the question that matters.

Sign up for Lake See the workflow ↓

An early product with a focused first connector: GitHub Repo, manual sync and read-only MCP access.

Lake

The foundation

One workspace.
Inspectable evidence.
Your questions.

01 / From connection to question

Start small.
Follow the evidence.

Lake brings selected repository records into connector-defined tables. It gives people and agents a common starting point for investigation—not an automatic answer to every business question.

  1. 01

    Connect a repository.

    Choose a GitHub repository, a unique table prefix and the datasets you need. Supply a fine-grained token with the required read permissions.

  2. 02

    Run a manual sync.

    Import selected records and inspect them in Data viewer. Check the sync result before treating the data as current.

  3. 03

    Give your agent context.

    Issue an MCP credential for a compatible client. Discover the tables, inspect the columns and retrieve relevant rows with read-only tools.

02 / Available source

GitHub, with
clear boundaries.

Connect a repository on github.com and choose the records your task needs. Slack and Google Drive are planned; scheduled sync is also planned.

Read the practical GitHub guide →
Issues
Build a review queue from open and closed issues. Pull requests are kept in their own table.
Pull requests
Inspect imported PR state, draft status and timestamps, then follow the source link for review context.
Commits
Explore commits from the repository's default branch. This is not a complete source-code archive.
Workflow runs
Investigate GitHub Actions run metadata. Jobs, logs, artifacts and third-party checks are not included.

03 / Illustrative investigation

What needs
a closer look?

A useful first question is small enough to verify. Prepare a queue of open pull requests for a person to review, rather than asking an agent to judge the health of a whole release.

Question → evidence → judgement

“Which open pull requests should we revisit at planning?”
  1. Discover. Inspect the imported pull request table and its columns.
  2. Retrieve. Filter open records; include draft status, timestamps and source links.
  3. Review. Open the evidence in GitHub. Imported state alone does not establish approval, urgency or readiness to merge.

04 / Before you connect

Know what
you're building on.

A shared foundation works best when its limits are visible.

Is this a live mirror of GitHub?

No. Sync is manual and paginated. A failed sync rolls back that run's imported rows. Previously imported rows remain after upstream deletion, force-pushes or dataset deselection. Check the last successful sync and verify important findings in GitHub.

Can I create saved views or run SQL?

Not in the current product. MCP supports structured single-table reads with column selection, filters, sorting and pagination. Saved views, custom columns, arbitrary SQL joins and aggregations are deferred.

Which agents can connect?

Clients must support Streamable HTTP with a custom bearer Authorization header. OpenCode is a documented example. Interactive MCP OAuth consent and discovery are not implemented; website sign-in is a separate flow.

What controls agent access?

Lake uses workspace-issued, named MCP credentials with thirty-day expiry and revocation. Tool calls recheck credential validity and issuer membership and write audit events. Current MCP tools are read-only. Choose the workspace deliberately; a credential name is not a custom table-level permission rule.

Can I connect Slack or Google Drive?

These are planned connectors, not available sources. Start with the supported GitHub Repo connector. There is no announced release date for additional sources or scheduled syncing.

Start with one real question

Your repository.
A clearer starting point.

Open Lake to sign in or create an account. Bring a repository you are authorised to connect and start with a manual sync.

Sign up for Lake