Grovedock MCP server
Commit Nov 11, 2026An open-source source control system for monorepos that allows AI agents to read, write, and land code via the Model Context Protocol.
Also known as kabirnarang39/pando
- Stars
- 1
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Nov 11, 2026
- License
- Apache-2.0
What is Grovedock MCP?
Grovedock is an open-source source control system for monorepos that allows AI agents to read, write, and land code via the Model Context Protocol.
Summary generated from the project's README
About Grovedock MCP server
Managing monorepos where hundreds of AI agents commit code simultaneously is what pando enables. This open-source source control system moves commits, rebases, and history management to the server to prevent the merge conflicts and slow clones typical of traditional git workflows. It is designed for developers and organizations that need to scale agentic workflows without the overhead of a merge queue.
Users can run the server locally using SQLite or deploy it at scale with Postgres, CockroachDB, and S3. The system can operate alongside existing hosts like GitHub or Bitbucket by using the pando mirror command to push changes back to them. For direct integration, agents connect via the Model Context Protocol using a URL and a token to read, write, and land code without needing to clone the entire repository history.
This tool provides specialized commands like pando cr for change requests and pando cat for fast file reading. It allows for extremely high throughput, capable of landing hundreds of changes per second, and offers sparse cloning that fetches only the specific file paths requested.
What you can do with it
- 1Scale agentic code contributionsAn AI agent can land hundreds of changes per second into a single branch without facing merge retries.
- 2Speed up CI workflowsA continuous integration system can read specific files directly from the server instead of performing a full repository checkout.
- 3Integrate with existing git hostsTeams can use this system alongside GitHub to allow agents to work in a high-performance environment while keeping their primary host updated.
- 4Perform rapid repository cloningDevelopers can initialize a workspace by fetching only the necessary file trees in a fraction of the time required by standard git.
Written from the project's README. Every use case is checked against a quote from it.
Automated checks
Ran Nov 16, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachablekabirnarang39/pando is public and reachablegithub.com/kabirnarang39/pando ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)Apache-2.0spdx.org/licenses/Apache-2.0.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-11-11github.com/kabirnarang39/pando ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubWhy Grovedock
git gives every clone the full history and lets every pusher race for the same branch.
That works for people. It breaks when agents and CI land tens of thousands of changes a
day into one repository: every job clones history it never reads, and every push to
main loses to someone else's and retries.
Google, Meta and Microsoft solved this with in-house systems that move history and commit creation to the server (prior art). Grovedock is that design as one open-source binary: SQLite on a laptop, or stateless replicas on Postgres and S3 behind caches.
Beside GitHub, or instead of it
You choose how far to go, and can change your mind at any point.
- Beside your git host. Agents land in Grovedock;
pando mirrorpushes every landed change to GitHub, GitLab or Bitbucket as ordinary commits, so CI, code search, issue links and reviewers keep working where they are. Nothing to migrate; stop the mirror to roll back. - Instead of it.
pando migrate https://github.com/acme/fleet --to SERVER/acme/fleetbrings the full history, open pull requests and issues across in one command. Grovedock hosts your repositories, people browse and review in its web console, and plain git, CI and editors connect throughgit-remote-pando. The history stays exportable to git at any time, so there's no lock-in either way.
Most teams start beside and move the busiest repositories over once the numbers are in.
Change requests, reviews and protected branches run on the server, in the console and from
the CLI (pando cr), for people and agents alike.
Highlights
-
Agents need a URL and a token, nothing else.
pando mcp --httpserves the whole server over the Model Context Protocol: any agent reads, writes and lands code, opens and reviews change requests and reports checks without cloning, limited to what its token allows. -
Landing without a merge queue. The server rebases concurrent changes that touch different files and moves the branch with one compare-and-swap per batch. 1,000 agents land 504 changes a second, with no client retries.
-
Clones without history. A clone fetches one tree, lazily, only for the paths you ask for: 0.46 s for git/git's 82,443 commits, against 19.1 s for
git clone. Or mount a revision on Linux, macOS or Windows and download files only when they are read. -
CI without checkouts.
pando cat main path/to/fileanswers in 9 ms. History, blame, diff and search are server APIs. -
Built for agents. A built-in MCP server, plus Python, TypeScript and Go SDKs.
-
Plain git still works.
git clone pando::https://…, then push and pull as usual. Import and export round-trip with identical trees, so you can leave any time. -
Ready for a company. SSO through OpenID Connect, per-directory read and write permissions, pre-land checks for CI and review, file locking for assets that cannot be merged, an audit log, backups with
pandod copy, TLS, Prometheus metrics. -
Runs on what you have. SQLite, Postgres or CockroachDB for metadata; disk, S3, GCS or Azure for objects. Every feature sits behind a flag.
Quick start
brew install kabirnarang39/tap/pando # or: curl -fsSL https://raw.githubusercontent.com/kabirnarang39/pando/main/install.sh | sh
pandod server --listen 127.0.0.1:7420 & # a server on SQLite
pando import-git ~/code/my-repo --server http://127.0.0.1:7420
pando clone http://127.0.0.1:7420 work && cd work
echo hello > hello.txt && pando commit -m "Say hello" && pando land
Everything else is in the documentation: Docker, Postgres and S3 setups in Deploy, every git equivalent in Coming from git, and every command and flag in the CLI and server references.
How it works
- The server creates commits. A client sends the files it changed, not a tree; the server builds the commit in time proportional to the change.
- Content is immutable and addressed by its BLAKE3 hash, so every cache tier can answer any request and verify it. Only branch pointers change.
- Landing is optimistic. Changes that touch different paths all land; real conflicts fail alone and leave the branch untouched.
Read the architecture and the design decisions.
Benchmarks
On git/git's history (82,443 commits), both systems on one machine:
| git | Grovedock | ||
|---|---|---|---|
| Full clone | 19.1 s | 0.46 s | 41× faster |
Sparse clone of Documentation/ |
9.5 s | 0.11 s | 90× faster |
| Read one file in CI, no clone | 149 ms | 9 ms | 16× faster |
log -n 50 -- builtin/add.c |
100 ms | 19 ms | 5× faster |
| 1,000 agents landing on one branch | 0.2 /s | 504 /s | no client retries |
| Server storage after compaction | 333 MB | 369 MB | 1.1× git's pack |
Where git wins: working offline with full history. Methods, raw numbers and limits are on the benchmark page.
Status
Grovedock is new and maintained by one person. The protocol (pando.v1), storage formats and
CLI follow semantic versioning, and CI runs the test suite on Linux, macOS and Windows.
The safest way to try it is next to git, through the git bridge. See the
changelog and roadmap.
Contributing
Issues and pull requests are welcome; start with CONTRIBUTING.md. Report security issues privately as described in SECURITY.md.
Licensed under Apache-2.0. The name is inspired by Pando, the aspen grove in Utah: some 47,000 trunks on one root. A grove is many agents on one history, and a dock is where their changes land.
Grovedock MCP questions
How do AI agents interact with Grovedock?
Agents can use the built-in MCP server to read, write, and land code, open and review change requests, and report checks without needing to clone the repository.
Can I use Grovedock alongside GitHub?
Yes, you can run it beside your existing git host by using the mirror feature to push landed changes to GitHub, GitLab, or Bitbucket.
Does Grovedock support standard git workflows?
Yes, plain git still works through git-remote-pando, allowing you to push and pull as usual with standard git commands.
What databases does Grovedock use for metadata?
Grovedock supports SQLite for local use, as well as Postgres and CockroachDB for production environments.