Is GitGhost only for developers?

No. Developers get serious git, review, pipeline, and security workflows. Non-technical builders can still describe work, follow progress, and review outcomes without learning every DevOps tool first.

Can I use my own coding agent?

Yes. The local agent bridge lets users connect favorite coding agents and lets GitGhost capture sessions, patches, checkpoints, and governed action requests.

Can teammates watch or pick up an agent session?

Teams can follow session activity, transcript checkpoints, branches, files touched, approvals, and review context from the project workspace. Shared live control will expand gradually, but the current goal is already clear: agent work should not vanish inside one private chat.

Will agents be allowed to change production code automatically?

Not by default. GitGhost is built around policies, approvals, merge requests, security gates, and audit trails so teams can decide what agents may do.

How is this different from GitHub or GitLab?

Traditional git hosts record code after work happens. GitGhost coordinates agent work as it happens: shared AI sessions, local agent capture, patches, pipelines, security scans, approvals, and evidence in one place.

What is available in beta?

The beta focuses on project hosting, AI chat, local agent session visibility, CLI setup, pipeline and security workflows, and approval/evidence. Hosted-agent and enterprise controls will expand gradually.

Multiplayer workspace for AI coding agents

Multiplayer AI for software teams.

GitGhost makes agent work multiplayer: teammates can follow sessions, review branches and diffs, approve risky actions, and inherit the context without chasing private transcripts.

Join the private beta See the product tour

Use GitGhost from your terminal

Install once, sign in, then connect local coding agents.

Claude Code Codex Gemini Cursor OpenCode

CLI guide

curl -fsSL https://gitghost.ai/install.sh | bash

Guided setup: gitghost-cli setup

Connect agent: gitghost-cli setup --agent <agent> --code <connect-code>

Agent work rail

Launch checkout MVP

review ready

Prompt

“Build the checkout flow, open a branch, run tests and scans, then prepare the merge request for review.”

Shared session opened

Teammate can follow

Branch and patch proposed

Evidence ready for review

Team approval required before merge

Watch agent work in project contextBring your favorite coding agentReview branches, diffs, and evidenceHand off with the full trail

Multiplayer agent workspace

See how private agent work becomes a shared team workflow.

Tour the surfaces beta users work inside: repo hosting, AI chat, local agent sessions, merge review, pipelines, security, and activity evidence in one connected workspace.

gitghost.ai/dashboard/projects/:id

product map

Project repository

Host the repo where agent work becomes reviewable.

The project page keeps the file tree, README, branch controls, commit context, clone actions, and project tabs connected to issues, merge requests, pipelines, security, and the agent sessions behind each change.

FilesCommitsBranches

FilesCommitsBranchesMerge requestsPipelinesSecurity

File tree

src/

src/App.jsx

src/index.css

package.json

README.md

main

Todo app

React app generated by a connected coding agent, committed on an AI branch, and ready for review.

Why GitGhost exists

AI coding is becoming team work. The git platform has to become multiplayer.

Agents can now work for hours across real repos, but most teams still manage them through private chats and scattered terminal logs. GitGhost gives the team one shared place to follow the session, review the diff, approve the risk, and keep the handoff trail.

For founders shipping with agents

Turn product ideas into issues, agent sessions, branches, and reviewable merge requests without becoming the only person who understands what the agent did.

AI chat for product requests
Guided project setup
Readable work evidence

For developers

Keep using Claude Code, Cursor, Codex, Gemini, OpenCode, or your own tools while GitGhost turns each session into shared context, diff evidence, approvals, and handoff history.

Local agent bridge
Patch proposals
Session checkpoints

For teams that cannot ship blind

Move faster without letting private agent chats bypass review. Policies, approvals, scans, and audit trails stay attached so teammates can inspect the work together.

Project policies
Security gates
Permission-aware workflows

Platform capabilities

One shared workspace from prompt to protected merge.

No more passing around read-only chat links, terminal screenshots, CI logs, and security notes just to let another teammate understand the work.

Code hosting for agent-generated work

Repos, branches, issues, merge requests, comments, and activity sit next to the agent sessions that produced each change.

Route work to the right agent

Create issues, inspect projects, propose patches, request reviews, prepare merge requests, and hand off tasks to local or hosted coding agents.

Pipelines without YAML first

Build and test projects with a pipeline engine designed for isolated runners, evidence, retries, logs, and future deployment gates.

Security from the first commit

Run code, dependency, secret, infrastructure, and web checks, then turn actionable findings into alerts or reviewed update requests.

Shared project memory with provenance

GitGhost remembers sessions, transcripts, checkpoints, approvals, and outcomes so a teammate can understand why a change exists.

Built for mixed teams

Technical and non-technical teammates can follow agent work together while developers keep the merge, policy, and review controls they expect.

Your agents, one shared control surface

Let people choose the agent. Let the team share the work.

The future is not one assistant. It is many agents, each with different strengths, working inside one team-readable workflow: connect, capture, follow, review, approve, merge.

Claude Code

Cursor

Codex

Gemini

OpenCode

Claude Code

Cursor

Codex

Gemini

OpenCode

Local session

feature/checkout-polish

connected

Shared transcript captured

Checkpoint saved for handoff

Patch proposal created

Approval request linked

What reviewers actually see

Every agent change gets a team-readable trail.

The point is not to let AI run wild. The point is to make agent work easy to inspect, discuss, approve, reproduce, and merge.

Shared agent session

What was asked, what the agent tried, what changed, and where the work paused for team review.

Prompt captured

Tool calls summarized

Checkpoint saved

Merge request context

Diff, issue link, reviewers, approvals, and generated summary stay attached to the change.

Files changed

Review comments

Approval state

Pipeline proof

Build logs, test results, artifacts, retry state, and failure reasons stay visible before merge.

Build passed

Tests passed

Artifacts attached

Security gate

Security findings, dependency risk, secrets, and policy decisions stay visible in the same review flow.

No leaked secrets

Dependencies checked

Policy evaluated

Trust by design

Multiplayer does not mean reckless.

GitGhost treats AI like a powerful teammate, not a magic admin account. Least privilege, visible approvals, security evidence, and traceable decisions stay in the workflow.

Agents get project-scoped access, not your whole account

Risky actions can require human approval before execution

Secrets and user credentials stay isolated from agent prompts

Security scans and update proposals are visible before merge

Every agent session keeps a trail of checkpoints and decisions

How it works

From private prompt to shared merge request.

Start with an idea, issue, or repo

Create a project, import a repository, or describe the work in the AI workspace.

Connect the agent you trust

Use GitGhost chat, connect a local coding agent, or run hosted agents when project policy allows it.

Make the work visible

See transcripts, checkpoints, changed files, requested actions, security results, and pipeline status while the work moves toward review.

Hand off, approve, and merge

Review the patch, run checks, keep the audit trail, and let teammates pick up the context without reconstructing five disconnected tools.

Why switch

AI work should not disappear inside private chats.

Old flow

One person prompts an agent, code appears somewhere else, CI runs in another tab, security lives in a backlog, and the team cannot inherit the context.

GitGhost flow

Agent sessions, branches, patches, approvals, scans, pipelines, and merge evidence stay connected so the next teammate can review or continue.

Questions

What beta users usually ask first.

Ready to make AI coding a team sport?

Join the beta, connect a repo or agent, and see how GitGhost turns private AI work into shared software delivery.

Create my beta account I already have access