CC-GG-bridgy

A tiny Cursor/VS Code companion extension that adds the one thing the official Claude Code extension doesn't have: a provider switch. One status-bar button flips your next Claude Code session between Anthropic (Claude subscription) and z.ai (GLM Coding Plan) — same extension UI, same session history, both directions. It also shows both providers' 5-hour usage side by side, right in the status bar.

⇄ Claude · C 28% · G 1%

It is a supervisor, not a fork: the official Claude Code extension stays untouched and does all the real work. Bridgy only decides which provider the CLI process talks to at spawn time, gates switching on "no answer pending", and walks you through resuming the conversation on the other side.

Validated live against Claude Code extension 2.1.220 (see DESIGN.md for the design record and the spike evidence).

Features

• Per-project provider toggle — one workspace can run GLM while every other window stays on Anthropic. Click the status-bar item to switch; the next new conversation in that project uses the new provider. No window reload, no settings churn.
• Dual usage readout — the status item shows both providers' 5-hour windows inline (C 28% · G 1%); the tooltip adds the weekly windows, reset times, and your GLM plan tier. The item takes the warning tint when the active provider's 5-hour window passes 80%.
• Busy gate — switching is gated while a response looks in-flight, so a toggle can't corrupt an open turn; a forced switch asks for confirmation.
• Native session handoff — transcripts are provider-agnostic and shared, so after a toggle you just start a new conversation and resume the previous session from Claude Code's own session list. Nothing is copied or proxied.
• Fail-open by design — on any doubt (missing state, unreadable config, provider endpoint down) the wrapper execs the real CLI untouched and the usage rows show the reason instead of erroring. Claude Code keeps working even if bridgy is misconfigured or deleted.

How it works

• Process-wrapper shim. Bridgy points claudeCode.claudeProcessWrapper (an official extension setting) at a small POSIX-sh shim. Every time the extension launches a Claude CLI process, the shim reads bridgy's per-project state and either execs the real binary clean (Anthropic) or with the GLM env injected (z.ai). Each new conversation is its own CLI process, which is why the toggle applies without a reload.

• Per-project state — ~/.config/cc-gg-bridgy/state.json maps workspace path → anthropic | glm (plus a default). The shim resolves the project from the spawned process's cwd, which is the workspace folder.

• GLM credentials — ~/.config/cc-gg-bridgy/glm.env (never committed; strict KEY=value lines, parsed not sourced). Per z.ai's Claude Code docs:

  ANTHROPIC_BASE_URL=https://api.z.ai/api/anthropic
  ANTHROPIC_AUTH_TOKEN=your-coding-plan-key
  ANTHROPIC_DEFAULT_OPUS_MODEL=glm-5.2[1m]
  ANTHROPIC_DEFAULT_SONNET_MODEL=glm-5.2[1m]
  ANTHROPIC_DEFAULT_HAIKU_MODEL=glm-4.7
  ANTHROPIC_SMALL_FAST_MODEL=glm-4.7
  

• Busy detection — bridgy finds the project's live session via Claude Code's session registry (~/.claude/sessions/<pid>.json), then classifies busy/idle from the transcript tail (including nested subagent activity). Long silent tool runs read as busy — the safe direction — and a 30-minute staleness escape stops a dead session from gating forever.

• Usage sources — the GLM side queries the Coding Plan quota endpoint with your key. The Claude side reads the rate_limits payload Claude Code itself hands to statusline scripts, teed to a file (see setup step 3) — no credential handling at all.

Install

Not on any marketplace (yet) — build from source:

git clone https://github.com/triartleet/cc-gg-bridgy
cd cc-gg-bridgy
pnpm install
pnpm build
pnpm dlx @vscode/vsce package --no-dependencies
# Cursor:
cursor --install-extension cc-gg-bridgy-*.vsix   # or: code --install-extension …

Requires the official Claude Code extension, macOS or Linux (the shim is POSIX sh — Windows would need a different wrapper), and pnpm/Node 20.

Setup

1. Configure the wrapper. On first activation bridgy offers to point claudeCode.claudeProcessWrapper at its shim (a stable copy under ~/.config/cc-gg-bridgy/, refreshed automatically on every activation). Decline and bridgy stays inert. This is a global setting — every window routes CLI spawns through the shim; on the Anthropic default the shim is a pure passthrough.

2. Add GLM credentials. Create ~/.config/cc-gg-bridgy/glm.env as shown above. No file → the GLM side simply reports itself unavailable.

3. (Optional) Feed the Claude usage readout. Claude Code only hands rate_limits to statusline scripts, so bridgy reads a tee of that payload. If you use a custom statusline, add this after it reads stdin (fail-safe — it can never break the status line itself):

   # after: input=$(cat)
   {
     bridgy_dir="$HOME/.config/cc-gg-bridgy"
     mkdir -p "$bridgy_dir" &&
       printf '%s' "$input" >"$bridgy_dir/statusline-last.json.tmp" &&
       mv -f "$bridgy_dir/statusline-last.json.tmp" "$bridgy_dir/statusline-last.json"
   } 2>/dev/null || true
   

   Without this, the Claude column shows why it's unavailable; everything else works.

Using it

Click ⇄ Claude / ⇄ GLM in the status bar to toggle the current project. Then start a new conversation (the toast offers it) — an open conversation keeps the provider it started on, by design. To continue a conversation on the other provider, resume it from Claude Code's session list; the transcript carries over natively.

Settings: ccGgBridgy.quietWindowMs — how long the transcript must be silent before a session counts as idle (default 2500 ms).

Debugging: touch ~/.config/cc-gg-bridgy/debug-on (or set CC_GG_BRIDGY_DEBUG=1 in the spawn env) makes the shim log its argv/cwd/provider decision to ~/.config/cc-gg-bridgy/debug.log (size-capped). CC_GG_BRIDGY_GLM_ENV overrides the glm.env path.

Limitations & caveats

• A fresh GLM conversation may open on the small/fast model slot (e.g. glm-4.7) depending on the panel's sticky model choice — check /model after switching. Bridgy deliberately never touches model choice.
• Claude usage freshness rides on terminal use. The extension UI doesn't run statusline scripts, so the Claude column updates when a terminal claude session makes a turn; past 30 minutes it's marked "as of HH:MM". The GLM column is always live.
• The busy heuristic reads safe, not perfect. Long silent tool executions and permission prompts read as busy; the forced-switch confirm covers the rest.
• Closed-source churn. Anthropic can change the wrapper setting or spawn path in any release (the extension auto-updates). The shim fails open, so the failure mode is "toggle silently does nothing", never a broken Claude Code — re-verify after major extension updates.
• Cosmetics: the extension UI may render Claude model names in some places while under GLM; the status bar is the truth surface for the provider.

Roadmap

• Button in the Claude panel's title bar beside the status-bar item.
• Auto-resume: open an integrated terminal running claude --resume <newest-session-id> under the target env.
• Per-session provider badges in the tooltip (from transcript model fields).
• Named env profiles instead of the fixed anthropic | glm pair — a universal per-project provider switch for any Anthropic-compatible endpoint.

Disclaimer

Not affiliated with Anthropic or Z.ai. Bridgy never proxies or intercepts provider traffic and never touches OAuth flows — it only sets an official extension setting and injects documented environment variables, so each provider is consumed exactly as its subscription intends. Claude Code is a product of Anthropic, PBC; use of each provider is governed by its own terms.

License

MIT