copilot-bridge 5.2.2

HTTP bridge that exposes VS Code + GitHub Copilot Chat to external Python scripts and agents

Copilot Bridge

A local HTTP bridge that exposes VS Code and GitHub Copilot Chat to any external process — Python scripts, agents, CLIs, or other tools.

External process
    ↕  HTTP (localhost:5150)
copilot-bridge-extension   ← VS Code extension (the server)
    ↕  VS Code API
GitHub Copilot Chat

Components

Path	Purpose
copilot-bridge-extension/	VS Code extension — starts the HTTP server inside VS Code
copilot-bridge-dist/	Python client library (CopilotBridge, CopilotAgent)

Quickstart

1 — Install the VS Code extension

From GitHub Releases (no build required):

1. Download copilot-bridge.vsix from the latest release
2. In VS Code: Ctrl+Shift+P → "Extensions: Install from VSIX..." → select the file

From source:

cd copilot-bridge-extension
npm install
npm run package          # produces copilot-bridge-x.x.x.vsix
code --install-extension copilot-bridge-*.vsix

The extension auto-starts an HTTP server on localhost:5150 when VS Code opens.

2 — Install the Python client

pip install copilot-bridge

From source:

cd copilot-bridge-dist
pip install .

3 — Use it

from copilot_bridge import CopilotBridge, CopilotAgent

# One-shot question
client = CopilotBridge()
print(client.ask("Explain this function in one sentence"))

# Stateful chat
client.set_system_prompt("You are a terse code reviewer.")
reply = client.chat("Review my PR diff")

# Agentic loop — reads/writes files, runs commands, etc.
agent = CopilotAgent()
agent.run("Refactor src/auth.py to use async/await")

CLI:

python -m copilot_bridge              # interactive chat
python -m copilot_bridge agent        # interactive agent
python -m copilot_bridge ask "Hello"  # one-shot
python -m copilot_bridge status       # connection check

API overview

CopilotBridge exposes the full VS Code surface over HTTP:

Category	Methods
Chat	ask, chat, chat_stream, chat_with_image
Files	read_file, write_file, edit_file, multi_edit, delete_file, rename_file, copy_file, list_directory
Search	search_text, search_files, find_files, semantic_search
Code intelligence	search_symbols, find_definition, find_usages, hover, document_symbols, rename_symbol, call_hierarchy
Git	git_status, git_diff, git_log, git_branches, git_add, git_commit, git_push, git_pull, git_merge, git_checkout, git_stash
Diagnostics	get_diagnostics, get_errors
Editor	get_editor, open_file, insert_text, get_selection
Terminal	create_terminal, send_to_terminal, get_terminal_output, dispose_terminal
UI	notify_info, notify_warn, notify_error, prompt_input, prompt_quickpick
Workspace	get_workspace, get_workspace_index, get_workspace_files, get_related_files, get_import_graph, reindex

See copilot-bridge-dist/copilot_bridge.pyi for the full typed interface.

Authentication

The bridge uses an API key to protect all endpoints. The key is generated automatically the first time the VS Code extension starts.

Key location: ~/.copilot-bridge/config.json

{ "apiKey": "a3f8c2d19e4b7a6f..." }

The Python client reads this file automatically — no extra setup needed:

client = CopilotBridge()  # key loaded from ~/.copilot-bridge/config.json

To use a custom key (e.g. in CI or a container):

client = CopilotBridge(api_key="your-key-here")

Or set it via environment variable in your own wrapper:

import os
client = CopilotBridge(api_key=os.environ["COPILOT_BRIDGE_KEY"])

To rotate the key: delete ~/.copilot-bridge/config.json and restart VS Code. A new key is generated automatically.

Note: /health is intentionally unauthenticated so the client can auto-discover the port.

Configuration

In VS Code settings (Ctrl+,, search "Copilot Bridge"):

Setting	Default	Description
copilotBridge.port	5150	HTTP server port (auto-increments if busy)
copilotBridge.autoStart	true	Start server when VS Code opens

Performance

Measured on v5.2.0 — localhost, 20 LLM rounds, 5 concurrent workers, 100 requests, 0 errors:

Benchmark	Mean	p95	p99
Health (HTTP only)	7 ms	17 ms	18 ms
Echo (ext host, no LLM)	8 ms	19 ms	24 ms
Sequential ask()	1949 ms	2138 ms	2651 ms
Streaming TTFT	1958 ms	2829 ms	2831 ms
Concurrent (5 workers, 100 req)	2018 ms	2732 ms	3054 ms

The bridge adds < 1 ms overhead (proven by /echo isolation). The LLM accounts for 99.96% of total latency.
Throughput: 2.45 req/s · 0 errors on 100 concurrent requests.

→ Full benchmark methodology and results

Security & Privacy

• Localhost only — the HTTP server binds exclusively to 127.0.0.1 and is never exposed to the network or the internet.
• API key protected — every endpoint (except /health) requires a bearer token stored in ~/.copilot-bridge/config.json. No unauthenticated request reaches Copilot Chat.
• No telemetry — the extension and Python client collect no usage data and make no outbound connections of their own.
• No data leaves your machine — all traffic stays between your local processes. Copilot Chat itself operates under your existing GitHub Copilot subscription and its own privacy terms.
• Thin bridge, not an agent runtime — this project is a dumb pipe. It does not store conversation history, cache credentials, or manage secrets beyond the single API key file.

Requirements

• VS Code 1.90+
• GitHub Copilot Chat extension
• Python ≥ 3.10 (client library)
• Node.js ≥ 18 (to build the extension from source)

License

MIT — see copilot-bridge-extension/LICENSE.