Docker MCPDocker MCP configuration — env file, workspace, and transportView Docker MCP configuration

Run gitaiflow MCP in Docker

Run the gitaiflow MCP server in Docker for an isolated or shared deployment.

Requires repository access

gitaiflow is a private repository, so this path is for repo collaborators, not a general public install. If you don’t have access, use the AI-Native Binary or Claude Desktop Extension instead — or request access.

Request repository access

The container packages the MCP layer and runs the gitaiflow native binary. Requires access to the private gitaiflow repository.

Clone the gitaiflow repository (requires repo access)

$git clone https://gitlab.com/codefleet-labs/gitaiflow.git
$cd gitaiflow

2
Choose the parent workspace that contains your repositories
```
$export MCP_HOST_WORKSPACE="/Users/<user>/projects"
```

Build the MCP image

$docker build --no-cache -f Dockerfile.mcp -t gitaiflow-mcp:local .

Run streamable HTTP (workspace, config and binary mounted)

$docker rm -f gitaiflow-mcp 2>/dev/null || true
$docker run --rm \
$  --name gitaiflow-mcp \
$  -p 8080:8080 \
$  --env-file .mcp.env \
$  -e MCP_TRANSPORT=streamable-http \
$  -e MCP_WORKSPACE=/workspace \
$  -e MCP_ALLOWED_WORKSPACES=/workspace \
$  -e GITAIFLOW_BINARY=/app/bin/gitaiflow \
$  -v "$HOME/.gitaiflow:/root/.gitaiflow" \
$  -v "$MCP_HOST_WORKSPACE:/workspace" \
$  -v "/path/to/gitaiflow-linux-arm64:/app/bin/gitaiflow:ro" \
$  gitaiflow-mcp:local

5
Check health
```
$curl -i http://localhost:8080/health
```

Docker deployment

The commands above are the local verification setup. Cloud Run is the next deployment step — follow the gitaiflow MCP documentation rather than reusing the local example unchanged.

Docker MCP configuration

Docker packages the MCP integration while the gitaiflow AOT executable remains the execution engine. Pass MCP settings with --env-file .mcp.env or -e flags; explicit -e values win, so .mcp.env can keep native host paths.

AI settings still come from ~/.gitaiflow/config.env — mount $HOME/.gitaiflow at /root/.gitaiflow (writable) so gitaiflow can read them and persist local state.

Variables

MCP_TRANSPORT

Default: streamable-http
Notes: streamable-http (default) serves /mcp and /health on port 8080; stdio talks over stdin/stdout for local desktop clients and starts no HTTP listener.

MCP_API_TOKEN

Default: Optional
Notes: Bearer token for the streamable HTTP endpoint. Clients send it as Authorization: Bearer <token>. Keep it out of logs and screenshots.

MCP_WORKSPACE

Default: Set explicitly
Notes: Working parent workspace. Native: an absolute host path. In Docker: /workspace.

MCP_ALLOWED_WORKSPACES

Default: Set explicitly
Notes: Allow-list that authorizes repository roots. Set it alongside MCP_WORKSPACE (native: same absolute path; Docker: /workspace). The request path then selects a Git repository beneath it.

GITAIFLOW_BINARY

Default: Auto-resolved
Notes: Path to the gitaiflow executable the MCP server runs. In Docker set it to /app/bin/gitaiflow and mount a matching Linux binary there.

Name	Default	Notes
MCP_TRANSPORT	streamable-http	`streamable-http` (default) serves `/mcp` and `/health` on port 8080; stdio talks over `stdin/stdout` for local desktop clients and starts no HTTP listener.
MCP_API_TOKEN	Optional	Bearer token for the streamable HTTP endpoint. Clients send it as `Authorization: Bearer <token>`. Keep it out of logs and screenshots.
MCP_WORKSPACE	Set explicitly	Working parent workspace. Native: an absolute host path. In Docker: `/workspace`.
MCP_ALLOWED_WORKSPACES	Set explicitly	Allow-list that authorizes repository roots. Set it alongside `MCP_WORKSPACE` (native: same absolute path; Docker: `/workspace`). The request path then selects a Git repository beneath it.
GITAIFLOW_BINARY	Auto-resolved	Path to the gitaiflow executable the MCP server runs. In Docker set it to /app/bin/gitaiflow and mount a matching Linux binary there.

Mounts

$HOME/.gitaiflow

Container: /root/.gitaiflow
Notes: Holds config.env with your AI settings. Must be writable — gitaiflow stores local state such as usage.jsonl there.

Parent workspace containing your repositories

Container: /workspace
Notes: Not required to be a Git repository itself; each repository beneath it must have its own .git.

Matching Linux gitaiflow binary

Container: /app/bin/gitaiflow (read-only)
Notes: Mount it explicitly — do not assume it exists just because the container starts.

Host	Container	Notes
$HOME/.gitaiflow	/root/.gitaiflow	Holds config.env with your AI settings. Must be writable — gitaiflow stores local state such as usage.jsonl there.
Parent workspace containing your repositories	/workspace	Not required to be a Git repository itself; each repository beneath it must have its own .git.
Matching Linux gitaiflow binary	/app/bin/gitaiflow (read-only)	Mount it explicitly — do not assume it exists just because the container starts.

Do not mount $PWD blindly

If $PWD is your gitaiflow checkout, -v "$PWD:/workspace" mounts only that one repository and sibling repositories disappear. Define the parent workspace explicitly: export MCP_HOST_WORKSPACE="/Users/<user>/projects".

Workspaces and repository selection

The Docker mount at /workspace, together with MCP_ALLOWED_WORKSPACES, identifies the parent workspace. The tool path argument selects one Git repository inside it, and the selected directory must itself be a Git repository root.

Multi-repository layout

Multiple repositories

Mount the parent directory that contains all your repositories to /workspace — one -v flag, not one per repository. For example, -v "/Users/<user>/projects:/workspace" if projects/ holds paperclip/, VoiceStudio/, and issuetracker/, each with its own .git.
gitaiflow selects a specific repository by name using the tool’s path argument — a request naming paperclip resolves to /workspace/paperclip.

Verify the container

Health check (expects HTTP 200)

$curl -i http://localhost:8080/health

Go deeper

Docker MCP verification guide (docs)Docker MCP troubleshooting (docs)

DjangoPlay · gitaiflow