Skip to content

Use Claude Code with your ChatGPT, Kimi, Cursor or Grok subscription via a local Anthropic-compatible proxy

Rust

support [1m] model suffix for larger context windows

Raine Virta65939c5aa47f
diff --git a/README.md b/README.md
index 73c69cf..b699648 100644
--- a/README.md
+++ b/README.md
@@ -104,16 +104,16 @@ would 400 because no provider claims it, so set
# Codex
ANTHROPIC_BASE_URL=http://localhost:18765 \
ANTHROPIC_AUTH_TOKEN=unused \
-ANTHROPIC_MODEL=gpt-5.4 \
-ANTHROPIC_SMALL_FAST_MODEL=gpt-5.4-mini \
+ANTHROPIC_MODEL=gpt-5.4[1m] \
+ANTHROPIC_SMALL_FAST_MODEL=gpt-5.4-mini[1m] \
CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1 \
claude
# Kimi
ANTHROPIC_BASE_URL=http://localhost:18765 \
ANTHROPIC_AUTH_TOKEN=unused \
-ANTHROPIC_MODEL=kimi-for-coding \
-ANTHROPIC_SMALL_FAST_MODEL=kimi-for-coding \
+ANTHROPIC_MODEL=kimi-for-coding[1m] \
+ANTHROPIC_SMALL_FAST_MODEL=kimi-for-coding[1m] \
CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1 \
claude
```
@@ -125,52 +125,28 @@ Or set it persistently in `~/.claude/settings.json`:
"env": {
"ANTHROPIC_BASE_URL": "http://127.0.0.1:18765",
"ANTHROPIC_AUTH_TOKEN": "unused",
- "ANTHROPIC_MODEL": "gpt-5.4",
- "ANTHROPIC_SMALL_FAST_MODEL": "gpt-5.4-mini",
+ "ANTHROPIC_MODEL": "gpt-5.4[1m]",
+ "ANTHROPIC_SMALL_FAST_MODEL": "gpt-5.4-mini[1m]",
"CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC": 1
}
}
```
-### 5. Optional: disable Claude Code auto-compact
+### 5. Context window size
-Claude Code decides auto-compaction locally based on the model context window it
-thinks it has. If the upstream model supports a larger window than Claude Code
-assumes, it may compact earlier than necessary.
+Claude Code decides auto-compaction based on the model's context window. For
+unknown models (like the ones the proxy uses) it defaults to 200K tokens, which
+is smaller than what the upstream models actually support (GPT-5.4: 400K+,
+Kimi: 256K). This causes auto-compact to fire earlier than necessary.
-Disable only automatic compaction while keeping manual `/compact` available:
+The `[1m]` suffix on the model name (shown in the examples above) is a Claude
+Code convention that tells it to use a 1M-token context window instead. This
+raises the auto-compact threshold without disabling it entirely.
-```sh
-DISABLE_AUTO_COMPACT=1 \
-ANTHROPIC_BASE_URL=http://localhost:18765 \
-ANTHROPIC_AUTH_TOKEN=unused \
-ANTHROPIC_MODEL=gpt-5.4 \
-ANTHROPIC_SMALL_FAST_MODEL=gpt-5.4-mini \
-CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1 \
- claude
-```
-
-Or add it to `~/.claude/settings.json`:
-
-```json
-{
- "env": {
- "ANTHROPIC_BASE_URL": "http://127.0.0.1:18765",
- "ANTHROPIC_AUTH_TOKEN": "unused",
- "ANTHROPIC_MODEL": "gpt-5.4",
- "ANTHROPIC_SMALL_FAST_MODEL": "gpt-5.4-mini",
- "CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC": 1,
- "DISABLE_AUTO_COMPACT": 1
- }
-}
-```
-
-Tradeoffs:
-
-- Claude Code will stop proactively compacting before a turn.
-- Manual `/compact` still works.
-- If you let the session grow too far, you may hit prompt-too-long failures
- instead of a graceful auto-compact.
+If you'd rather disable auto-compact completely, set
+`DISABLE_AUTO_COMPACT=1` in your env or `~/.claude/settings.json`. Manual
+`/compact` still works, but you risk hitting real upstream limits before
+Claude Code can compact for you.
## Toggling between proxy and direct Anthropic
@@ -190,10 +166,9 @@ the default.
if [ -f "$HOME/.claude/claude-code-proxy-enabled" ]; then
export ANTHROPIC_BASE_URL="http://localhost:18765"
export ANTHROPIC_AUTH_TOKEN="unused"
- export ANTHROPIC_MODEL="gpt-5.4"
- export ANTHROPIC_SMALL_FAST_MODEL="gpt-5.4-mini"
+ export ANTHROPIC_MODEL="gpt-5.4[1m]"
+ export ANTHROPIC_SMALL_FAST_MODEL="gpt-5.4-mini[1m]"
export CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC="1"
- export DISABLE_AUTO_COMPACT=1
fi
exec "$HOME/.local/bin/claude" "$@"
diff --git a/src/cli.ts b/src/cli.ts
index f873bf1..3705186 100755
--- a/src/cli.ts
+++ b/src/cli.ts
@@ -38,8 +38,8 @@ async function main() {
console.log("Configure Claude Code (pick a model from above):")
console.log(` export ANTHROPIC_BASE_URL="http://localhost:${port}"`)
console.log(` export ANTHROPIC_AUTH_TOKEN="anything"`)
- console.log(` export ANTHROPIC_MODEL="kimi-for-coding" # or gpt-5.4, etc.`)
- console.log(` export ANTHROPIC_SMALL_FAST_MODEL="kimi-for-coding" # background / title-gen`)
+ console.log(` export ANTHROPIC_MODEL="kimi-for-coding[1m]" # or gpt-5.4[1m], etc.`)
+ console.log(` export ANTHROPIC_SMALL_FAST_MODEL="kimi-for-coding[1m]" # background / title-gen`)
console.log(` export CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC="1"`)
return
}
diff --git a/src/server.ts b/src/server.ts
index 54b56c6..32c5c1b 100644
--- a/src/server.ts
+++ b/src/server.ts
@@ -100,6 +100,15 @@ function buildCtx(req: Request, reqId: string, providerName: string): RequestCon
}
}
+// Claude Code uses a [1m] suffix convention (e.g. "gpt-5.4[1m]") to
+// signal that the model should be treated as having a 1M-token context
+// window. Claude Code normalizes this away before sending requests to
+// the API, but we strip it here too as defense-in-depth in case a
+// future version or a different client includes it.
+function normalizeIncomingModel(model: string): string {
+ return model.replace(/\[1m\]$/i, "")
+}
+
function routeProvider(body: AnthropicRequest, reqId: string): Provider | Response {
if (!body.model) {
return jsonError(
@@ -108,6 +117,7 @@ function routeProvider(body: AnthropicRequest, reqId: string): Provider | Respon
`Missing "model" in request body. ${knownModelsMessage()}`,
)
}
+ body.model = normalizeIncomingModel(body.model)
const provider = providerForModel(body.model)
if (!provider) {
rootLog.warn("unknown model", { reqId, model: body.model })