Skip to content

Latest commit

 

History

History
168 lines (120 loc) · 3.93 KB

File metadata and controls

168 lines (120 loc) · 3.93 KB

SloplessCode Client Setup

This guide is for a client machine that does not run the SloplessCode server. It only connects to an already running SloplessCode server over the network.

Requirements

  • Python 3.9+
  • httpx (pip install httpx)
  • Claude Code
  • The mcp/server.py file from this repository, or a full clone of the repository

Qdrant, Ollama, Docker, and FastAPI are not required on the client machine.

1. Check Server Reachability

Replace <SERVER_IP> with the server machine address:

curl http://<SERVER_IP>:8000/api/v1/health

If the server is unreachable, check that:

  • docker compose up -d is running on the server;
  • port 8000 is open in the firewall;
  • the server listens on 0.0.0.0, not only on 127.0.0.1;
  • an API_KEY requirement is not missing from the client configuration.

2. Get The MCP Client File

Option A: copy only the stdio bridge file:

mkdir <CLIENT_HOME>
# Copy mcp\server.py from the repository or from the server by scp/sftp.
mkdir -p ~/sloplesscode-client
scp user@<SERVER_IP>:/path/to/sloplesscode/mcp/server.py ~/sloplesscode-client/

Option B: clone the full repository:

git clone https://github.com/Utundry/sloplesscode.git sloplesscode-client

3. Install The Python Dependency

For a simple client-only setup:

pip install httpx

Or use a virtual environment:

python -m venv .venv
.\.venv\Scripts\activate
pip install httpx
python3 -m venv .venv
source .venv/bin/activate
pip install httpx

4. Register SloplessCode In Claude Code

Find the claude executable:

# Windows VS Code extension path, adjust the wildcard if needed.
$env:USERPROFILE\.vscode\extensions\anthropic.claude-code-*\resources\native-binary\claude.exe
which claude

Register the client bridge.

Windows:

claude mcp add -s user `
  -e "MEMORY_SERVER_URL=http://<SERVER_IP>:8000" `
  sloplesscode `
  -- "<CLIENT_HOME>\.venv\Scripts\python.exe" "<CLIENT_HOME>\server.py"

Linux or macOS:

claude mcp add -s user \
  -e "MEMORY_SERVER_URL=http://<SERVER_IP>:8000" \
  sloplesscode \
  -- ~/sloplesscode-client/.venv/bin/python ~/sloplesscode-client/server.py

If the server requires an API key, add the expected environment variable or client header according to your MCP client configuration. For SSE clients, send X-Api-Key: <your key>.

Verify the registration:

claude mcp list

5. Optional Skills

If your client uses Claude skills, copy the repository skill files:

Windows:

mkdir "$env:USERPROFILE\.claude\skills\remember"
mkdir "$env:USERPROFILE\.claude\skills\recall"
copy skills\remember\SKILL.md "$env:USERPROFILE\.claude\skills\remember\SKILL.md"
copy skills\recall\SKILL.md "$env:USERPROFILE\.claude\skills\recall\SKILL.md"

Linux or macOS:

mkdir -p ~/.claude/skills/remember ~/.claude/skills/recall
cp skills/remember/SKILL.md ~/.claude/skills/remember/
cp skills/recall/SKILL.md ~/.claude/skills/recall/

Restart VS Code after changing Claude Code MCP or skill configuration.

6. Smoke Test

Use a small synthetic memory:

/remember --agent mypc --type fact Client machine is connected to the memory server
/recall client connection to memory server

Multiple Clients

Each client can use its own agent_id or share a team-level agent_id:

/remember --agent alice Prefers vim
/remember --agent team Deploy every Friday at 18:00
/recall --agent team deployment process

Topology

Client A (Windows)              Client B (Linux)
  Claude Code                     Claude Code
  mcp/server.py  ------------>    mcp/server.py
        |                               |
        +---------- HTTP ---------------+
                          |
                   <SERVER_IP>:8000
                   SloplessCode API
                          |
                       Qdrant
                 optional LLM providers