Git Repositories

• Store code and assets produced by coding agents, one repository per generated project
• Give agents a durable Git history with branches, merges, and activity attribution
• Create repositories at product scale from your control plane with the SDKs
• Mount a repository as a live directory in a sandbox, without cloning it first
• Track durable workspace checkpoints, snapshots, branch activity, and mount liveness from the control plane

​

Quickstart

curl -fsSL https://tensorlake.ai/install | sh
tl login

$ tl git create agent-outputs --default-branch main
created https://git.tensorlake.ai/project_9f3c2a1b/agent-outputs

$ pip install tensorlake
$ export TENSORLAKE_API_KEY=tlk_...

from tensorlake import RepositoryClient

with RepositoryClient.from_env() as repos:
    repo = repos.create("agent-outputs", default_branch="main")
    print(repo.url)

$ npm install tensorlake
$ export TENSORLAKE_API_KEY=tlk_...

import { RepositoryClient } from "tensorlake";

const repos = await RepositoryClient.fromEnv();
const repo = await repos.create("agent-outputs", {
  defaultBranch: "main",
});

console.log(repo.url);

$ TOKEN=$(tl git token --repo agent-outputs --json | jq -r .token)
$ git clone https://t:$TOKEN@git.tensorlake.ai/project_9f3c2a1b/agent-outputs
$ cd agent-outputs

$ echo "# Agent Outputs" > README.md
$ git add README.md
$ git commit -m "initial app"
$ git push origin main
$ git log --oneline -1
4f8c2a1 initial app

$ mkdir -p initial-app
$ echo "# Agent Outputs" > initial-app/README.md

from tensorlake import RepositoryClient

with RepositoryClient.from_env() as repos:
    report = repos.push_worktree(
        "agent-outputs",
        root="./initial-app",
        branch="main",
        message="initial app",
    )
    print(report.commit)

$ mkdir -p initial-app
$ echo "# Agent Outputs" > initial-app/README.md

import { RepositoryClient } from "tensorlake";

const repos = await RepositoryClient.fromEnv();
const report = await repos.pushWorktree("agent-outputs", {
  path: "./initial-app",
  branch: "main",
  message: "initial app",
});

console.log(report.commit);

$ tl git mount agent-outputs /work
Mounted writable agent-outputs at /work (workspace 3f9a2b7e1c4d activates on the first WAL checkpoint)
$ ls /work
README.md

$ printf "\n## Parser Notes\n" >> /work/README.md
$ tl git snapshot /work -m "add notes"
Snapshot 8b21f6a9c3d5e7f1a2b4c6d8e0f3a5b7c9d1e3f5 (1 file(s), 1 of 1 chunks uploaded)
$ tl git promote /work main
Promoted workspace 3f9a2b7e1c4d -> main at 1c4d9a2f7e5b3d8c6a4f2e0d9b7c5a3f1e8d6b4c (squashed)
$ git -C agent-outputs fetch origin main
$ git -C agent-outputs log --oneline -2 origin/main
1c4d9a2 add notes
4f8c2a1 initial app

​

Mental Model

1. A repository stores the durable Git history.
2. A mount gives a sandbox an ephemeral directory backed by the repository, no clone required. A mount is writable by default; add --ro for a stateless read-only view of a branch, a pinned commit, or a subtree of a monorepo.
3. The shared local journal records writes crash-safely. Autosave publishes those writes to a durable private workspace WAL without creating a commit or moving a branch.
4. A workspace is created lazily on the first remote WAL checkpoint and holds an agent’s isolated state.
5. A snapshot materializes the current WAL as a Git commit on the workspace.
6. Promotion publishes a workspace snapshot to a branch so future mounts and Git users can use that version. Rebase replays a workspace onto a moved branch, server-side.
7. The control plane can observe mount liveness and durable workspace operations. The unsealed local edit tail stays in the sandbox until autosave.

​

Git or the tl CLI?

• You cloned (git clone) → use ordinary Git. A clone is a real checkout with a .git directory; Tensorlake is a normal remote. Commit, rebase, merge, and push exactly as you would anywhere.
• You mounted (tl git mount) → use the tl git verbs. A mount has no .git and Git commands do not run inside it; the verbs are the mount’s interface, and each maps onto a Git habit:

  Git habit on a clone	Equivalent on a mount
  git commit	tl git snapshot
  git fetch / switch a pristine view	tl git sync
  git rebase origin/main	tl git rebase
  git push	tl git promote
  git status / git log	tl git status / tl git log

• You have no files at all (a control plane, CI, an SDK caller) → use tl git merge and the SDKs. Merges, preflights, and conflict queries run server-side without any checkout.

​

What Tensorlake Provides

• Plain Git repositories: clone, branch, commit, merge, push, and fetch over Git smart HTTP.
• Server-side merges and rebases: merge branches, preflight conflicts, rebase workspaces, and query structured conflict records without a checkout.
• Repository mounts: instant read-only views of any branch, commit, or subtree — and isolated workspaces with private autosave WAL plus explicit snapshot, promote, and rebase when agents need to write.
• Fleet observability: durable workspace checkpoints, snapshots, operations, and mount liveness are available to the control plane, with actor attribution in activity history.
• SDKs: create repositories, push worktrees, and merge branches from Python or TypeScript.

​

Where To Go Next