Context Engine Uploader

Install

• Install from the VS Code Marketplace (search for "Context Engine Uploader" or publisher context-engine). You can also install it directly from the Extensions view in VS Code.

Features

• Runs a force sync (Index Codebase) followed by watch mode to keep a remote Context Engine instance in sync with your workspace.
• Auto-detects the first workspace folder as the default target path, storing it in workspace settings so the extension is portable.
• Provides commands and a status-bar button:
  • Context Engine Uploader: Index Codebase – force sync + watch with spinner feedback.
  • Context Engine Uploader: Start/Stop/Restart for manual lifecycle control.
• Streams detailed logs into the Context Engine Upload output channel for visibility into both force sync and watch phases.
• Status bar shows current state (indexing spinner, purple watching state) so you always know if uploads are active.

Configuration

• Run On Startup auto-triggers force sync + watch after VS Code finishes loading.
• Python Path, Endpoint, Extra Force Args, Extra Watch Args, and Interval Seconds can be tuned via standard VS Code settings.
• Target Path is auto-filled from the workspace but can be overridden if you need to upload a different folder.
• Python dependencies: the extension runs the standalone upload client via your configured pythonPath. Ensure the interpreter has requests, urllib3, and charset_normalizer installed. Run python3 -m pip install requests urllib3 charset_normalizer (or replace python3 with your configured path) before starting the uploader.
• Path mapping: Host Root + Container Root control how local paths are rewritten before reaching the remote service. By default the host root mirrors your Target Path and the container root is /work, which keeps Windows paths working without extra config.
• Prompt+ decoder: set Context Engine Uploader: Decoder Url (default http://localhost:8081, auto-appends /completion) to point at your local llama.cpp decoder. For Ollama, set it to http://localhost:11434/api/chat. Turn on Use Gpu Decoder to set USE_GPU_DECODER=1 so ctx.py prefers the GPU llama.cpp sidecar. Prompt+ automatically runs the bundled scripts/ctx.py when an embedded copy is available, falling back to the workspace version if not.
• Claude/Windsurf MCP config:
  • MCP Indexer Url and MCP Memory Url control the URLs written into the project-local .mcp.json (Claude) and Windsurf mcp_config.json when you run the Write MCP Config command. These URLs are used literally (e.g. http://localhost:8001/sse or http://localhost:8003/mcp).
  • MCP Server Mode (contextEngineUploader.mcpServerMode) controls what servers are written:
    • bridge: write a single context-engine server that talks to the ctxce MCP bridge.
    • direct: write two servers, qdrant-indexer and memory, that talk directly to the configured URLs.
  • MCP Transport Mode (contextEngineUploader.mcpTransportMode) controls how those servers talk:
    • sse-remote (default): use stdio MCP processes behind an SSE tunnel (bridge-stdio / direct-sse).
    • http: use HTTP MCP endpoints directly (bridge-http / direct-http).
  • Common combinations:
    • bridge + sse-remote → bridge-stdio: write a single context-engine stdio server that runs ctxce mcp-serve behind an SSE tunnel.
    • bridge + http → bridge-http: write a single context-engine HTTP server that points at the local ctxce mcp-http-serve URL (e.g. http://127.0.0.1:30810/mcp).
    • direct + http → direct-http: write separate HTTP servers for the indexer and memory MCP backends.
• MCP config on startup:
  • contextEngineUploader.autoWriteMcpConfigOnStartup (default false) controls whether the extension automatically runs the same logic as Write MCP Config on activation. When enabled, it refreshes .mcp.json, Windsurf mcp_config.json, and the Claude hook (.claude/settings.local.json) to match your current settings and the installed extension version. If scaffoldCtxConfig is also true, this startup path will additionally scaffold/update ctx_config.json and .env as described below.
• CTX + GLM settings:
  • contextEngineUploader.ctxIndexerUrl is copied into .env (as MCP_INDEXER_URL) so the embedded ctx.py knows which MCP indexer to call when enhancing prompts.
  • contextEngineUploader.glmApiKey, glmApiBase, and glmModel are used when scaffolding ctx_config.json/.env to pre-fill GLM decoder options. Existing non-placeholder values are preserved, so you can override them in the files at any time.
• Git history upload settings:
  • contextEngineUploader.gitMaxCommits controls REMOTE_UPLOAD_GIT_MAX_COMMITS, bounding how many commits the upload client includes per bundle (set to 0 to disable git history).
  • contextEngineUploader.gitSince controls REMOTE_UPLOAD_GIT_SINCE, letting you constrain the git log window (e.g. 2 years ago or 2023-01-01).
• Context scaffolding:
  • contextEngineUploader.scaffoldCtxConfig (default true) controls whether the extension keeps a minimal ctx_config.json + .env in sync with your workspace. When enabled, running Write MCP Config or Write CTX Config will reuse the workspace’s existing files (if present) and only backfill placeholder or missing values from the bundled env.example plus the inferred collection name. Existing custom values are preserved.
  • The scaffolder also enforces CTX defaults (e.g., MULTI_REPO_MODE=1, REFRAG_RUNTIME=glm, REFRAG_DECODER=1) so the embedded ctx.py is ready for remote uploads, regardless of the “Use GLM Decoder” toggle.
  • contextEngineUploader.surfaceQdrantCollectionHint gates whether the Claude hook adds a hint line with the Qdrant collection ID when ctx is enhancing prompts. This setting is also respected when the extension writes .claude/settings.local.json.

MCP bridge (ctx-mcp-bridge) & MCP config lifecycle

• The MCP bridge (@context-engine-bridge/context-engine-mcp-bridge, CLI ctxce) is a small local MCP server that fans out to two upstream MCP services: the Qdrant indexer and the memory/search backend. The VS Code extension can drive it in two ways:
  • Bridge stdio (bridge-stdio) – a stdio MCP server (ctxce mcp-serve) wrapped behind an SSE tunnel.
  • Bridge HTTP (bridge-http) – an HTTP MCP server (ctxce mcp-http-serve) listening on http://127.0.0.1:<port>/mcp.
• Why use the bridge instead of two direct MCP entries?
  • Single server entry: IDEs only need to register one MCP server (context-engine) instead of juggling separate qdrant-indexer and memory entries, avoiding coordination mistakes.
  • Shared session defaults: the bridge loads ctx_config.json and injects collection name, repo metadata, and any other ctx defaults so every IDE window talks to the right collection without hand-editing .mcp.json.
  • Per-user credential isolation: each IDE maintains its own MCP session while the bridge multiplexes upstream calls. When you enable backend auth (via CTXCE_AUTH_ENABLED and ctxce auth ... sessions), uploads and MCP calls are gated by per-user sessions, so multiple IDEs can share the same stack while still having isolated access control and preferences. See Optional auth with the MCP bridge (PoC) below for details.
  • Flexible transport: stdio mode works everywhere (even when HTTP ports aren’t reachable), while HTTP mode keeps Claude/Windsurf happy when they want direct URLs; the extension automatically writes the right flavor.
  • Centralized logging & health: when the bridge process runs once per workspace you get a single stream of logs (Context Engine Upload output) and a single port to probe for health checks instead of multiple MCP child processes per IDE.
• When you run Write MCP Config, the extension:
  • Writes .mcp.json in the workspace for Claude Code.
  • Optionally writes Windsurf’s mcp_config.json (when mcpWindsurfEnabled=true).
  • Optionally scaffolds ctx_config.json + .env (when scaffoldCtxConfig=true).
• The effective wiring mode is determined by the two MCP settings:
  • mcpServerMode = bridge, mcpTransportMode = sse-remote → bridge-stdio.
  • mcpServerMode = bridge, mcpTransportMode = http → bridge-http.
  • mcpServerMode = direct, mcpTransportMode = sse-remote → direct-sse (two stdio mcp-remote servers).
  • mcpServerMode = direct, mcpTransportMode = http → direct-http (two HTTP servers, no bridge).
• In bridge-stdio, the configs run the ctxce mcp-serve CLI via npx (for example, npx @context-engine-bridge/context-engine-mcp-bridge ctxce mcp-serve), passing the workspace path (auto-detected from the uploader target path) plus --indexer-url and --memory-url derived from the MCP settings.
• In bridge-http, the extension can also manage the bridge process:
  • autoStartMcpBridge=true and mcpServerMode='bridge' with mcpTransportMode='http' → the extension starts ctxce mcp-http-serve in the background for the active workspace using mcpBridgePort.
  • The resulting HTTP URL (http://127.0.0.1:<mcpBridgePort>/mcp) is written into .mcp.json and Windsurf’s mcp_config.json as the context-engine server URL.
  • In stdio or direct modes, the HTTP bridge is not auto-started; only the explicit Start MCP HTTP Bridge command will launch it.
• Bridge settings are workspace-scoped, so different workspaces can choose different modes and ports (e.g., one workspace using stdio bridge, another using HTTP bridge on a different port).

Optional auth with the MCP bridge (PoC)

Auth is off by default and fully opt-in. When enabled, the MCP indexer and memory servers expect a valid session id (issued by the backend) on protected tools. The bridge CLI (ctxce auth ...) is the primary way to obtain and cache that session.

High-level steps:

• Enable auth on the remote stack (e.g. dev-remote compose):
  • Set CTXCE_AUTH_ENABLED=1 in the upload/indexer environment.
  • Optionally set CTXCE_AUTH_SHARED_TOKEN for token-based login.
  • Optional: set CTXCE_AUTH_ADMIN_TOKEN for creating additional users via /auth/users.
  • Optional (dev-only): set CTXCE_AUTH_ALLOW_OPEN_TOKEN_LOGIN=1 only if you want /auth/login to issue sessions even when CTXCE_AUTH_SHARED_TOKEN is unset. By default (0/unset), token-based login is disabled when no shared token is configured.
• Point the bridge at the auth backend:
  • In your local shell (where you run the ctxce CLI via npx), you can either:
    • Explicitly set CTXCE_AUTH_BACKEND_URL to the upload service URL (e.g. http://localhost:8004), or
    • Let the CLI discover the backend URL in this order when you run ctxce auth login: --backend-url / --auth-url → CTXCE_AUTH_BACKEND_URL → any stored auth entry → the uploader's upload_endpoint (CTXCE_UPLOAD_ENDPOINT / UPLOAD_ENDPOINT) → http://localhost:8004.

Token-based login:

export CTXCE_AUTH_BACKEND_URL=http://localhost:8004   # optional when using this extension
export CTXCE_AUTH_TOKEN=change-me-dev-token   # must match CTXCE_AUTH_SHARED_TOKEN in the stack

# Obtain a session and cache it under ~/.ctxce/auth.json
npx @context-engine-bridge/context-engine-mcp-bridge ctxce auth login

# Check status (optional)
npx @context-engine-bridge/context-engine-mcp-bridge ctxce auth status

Username/password login (when you have real users):

• First, create the initial user (once) via /auth/users while the auth DB is empty (no admin token required). This is typically done with a small script or curl call against the upload service, for example:

  curl -X POST http://localhost:8004/auth/users \
    -H "Content-Type: application/json" \
    -d '{"username":"you@example.com","password":"your-password"}'
  

• Then login via the bridge:

npx @context-engine-bridge/context-engine-mcp-bridge ctxce auth login \
  --username you@example.com \
  --password 'your-password'

In both modes, the bridge stores the returned session_id keyed by CTXCE_AUTH_BACKEND_URL and automatically injects it into all MCP tool calls as the session field. Once you have at least one entry in ~/.ctxce/auth.json, the MCP bridge used by the extension will discover and reuse that session automatically; MCP configs do not need to set CTXCE_AUTH_BACKEND_URL in their env blocks. If CTXCE_AUTH_ENABLED is off on the backend, these auth settings are ignored and the bridge behaves exactly as before.

Session lifetime:

• By default, issued sessions do not expire (CTXCE_AUTH_SESSION_TTL_SECONDS defaults to 0).
• Operators who want expiry can set CTXCE_AUTH_SESSION_TTL_SECONDS (in seconds) on the backend services. Values > 0 enable a sliding window: active sessions are refreshed when validated; values <= 0 disable expiry.

Workspace-level ctx integration

• The VSIX bundles an env.example template plus the ctx hook/CLI so you can dogfood the workflow without copying files manually.
• When scaffolding is enabled (see above), running the Context Engine Uploader: Write CTX Config (ctx_config.json/.env) command will:
  • Infer the collection name from the standalone upload client (--show-mapping).
  • Create or update ctx_config.json with that collection and sensible defaults (GLM runtime, default_mode, require_context, etc.).
  • Create or update .env from the bundled template, ensuring CTX-critical values such as MULTI_REPO_MODE=1, REFRAG_RUNTIME=glm, and REFRAG_DECODER=1 are set. Non-placeholder values (e.g., a real GLM_API_KEY) are left alone.
• You still own the files: if you need a custom value, edit .env or ctx_config.json directly. The scaffolder only touches keys that are missing, empty, or obviously placeholders.
• The Claude hook + ctx prompt enhancement is currently wired for Linux/dev-remote environments only. On other platforms, MCP config and uploading still work, but the automatic prompt rewrite hook is disabled.

Commands

• Command Palette → “Context Engine Uploader” exposes Start/Stop/Restart/Index Codebase and Prompt+ (unicorn) rewrite commands.
• Status-bar button (Index Codebase) mirrors Start/Stop/Restart/Index status, while the Prompt+ status button runs the ctx rewrite command on the current selection.
• Context Engine Uploader: Write MCP Config (.mcp.json) writes or updates a project-local .mcp.json (plus Windsurf mcp_config.json when enabled) using the currently selected bridge/direct + transport modes. If bridge-http is required and not yet running, the extension starts ctxce mcp-http-serve before writing configs.
• Context Engine Uploader: Write CTX Config (ctx_config.json/.env) scaffolds the ctx config + env files as described above. This command runs automatically after Write MCP Config if scaffolding is enabled, but it is also exposed in the Command Palette for manual use.
• Context Engine Uploader: Upload Git History (force sync bundle) triggers a one-off force sync using the configured git history settings, producing a bundle that includes a metadata/git_history.json manifest for remote lineage ingestion.
• Context Engine Uploader: Start MCP HTTP Bridge launches ctxce mcp-http-serve using the workspace’s resolved target path, MCP URLs, and configured mcpBridgePort. Use this when you want to run the HTTP bridge manually (e.g., testing unpublished builds or sharing a port across IDEs).
• Context Engine Uploader: Stop MCP HTTP Bridge gracefully terminates a running HTTP bridge process.

Logs

Open View → Output → Context Engine Upload to see the remote uploader’s stdout/stderr, including any errors from the Python client.