Skip to content

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

Rust

docs: correct codex context guidance

Raine Virta23f7fdb76599
diff --git a/README.md b/README.md
index c434f0f..8d27f9e 100644
--- a/README.md
+++ b/README.md
@@ -109,8 +109,8 @@ 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[1m] \
-ANTHROPIC_SMALL_FAST_MODEL=gpt-5.4-mini[1m] \
+ANTHROPIC_MODEL=gpt-5.5 \
+ANTHROPIC_SMALL_FAST_MODEL=gpt-5.4-mini \
CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1 \
CLAUDE_CODE_DISABLE_NONSTREAMING_FALLBACK=1 \
claude
@@ -138,8 +138,8 @@ 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[1m]",
- "ANTHROPIC_SMALL_FAST_MODEL": "gpt-5.4-mini[1m]",
+ "ANTHROPIC_MODEL": "gpt-5.5",
+ "ANTHROPIC_SMALL_FAST_MODEL": "gpt-5.4-mini",
"CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC": 1,
"CLAUDE_CODE_DISABLE_NONSTREAMING_FALLBACK": 1
}
@@ -149,13 +149,19 @@ Or set it persistently in `~/.claude/settings.json`:
### 5. Context window size
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.
+unknown models, Claude Code uses its own fallback context size. Do not append
+`[1m]` to Codex model names unless the upstream model really supports a 1M-token
+window. Current official Codex metadata reports `gpt-5.5` with a 272K-token
+window, so `gpt-5.5[1m]` makes Claude Code wait too long before compacting and
+can cause upstream context-window errors.
-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.
+Use the plain Codex model name for Codex models, such as `gpt-5.5` or
+`gpt-5.4-mini`. The proxy still strips a trailing `[1m]` for compatibility, but
+that suffix only affects Claude Code's local compaction threshold. It does not
+increase Codex's upstream context window.
+
+Kimi can still use `kimi-for-coding[1m]` if you want Claude Code to use a larger
+local compaction threshold for that provider.
If you'd rather disable auto-compact completely, set
`DISABLE_AUTO_COMPACT=1` in your env or `~/.claude/settings.json`. Manual
@@ -180,8 +186,8 @@ the default.
if [ -f "$HOME/.claude/claude-code-proxy-enabled" ]; then
export ANTHROPIC_BASE_URL="http://localhost:18765"
export ANTHROPIC_AUTH_TOKEN="unused"
- export ANTHROPIC_MODEL="gpt-5.4[1m]"
- export ANTHROPIC_SMALL_FAST_MODEL="gpt-5.4-mini[1m]"
+ export ANTHROPIC_MODEL="gpt-5.5"
+ export ANTHROPIC_SMALL_FAST_MODEL="gpt-5.4-mini"
export CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC="1"
export CLAUDE_CODE_DISABLE_NONSTREAMING_FALLBACK="1"
fi
@@ -223,8 +229,8 @@ Upstream: `https://chatgpt.com/backend-api/codex/responses` (Responses API).
Set `ANTHROPIC_MODEL` to a model your ChatGPT subscription is allowed to use.
Append `-fast` to a Codex model name to request Codex fast mode for that request
-without restarting the proxy. For example, `gpt-5.4-fast[1m]` is sent upstream as
-model `gpt-5.4` with `service_tier: "priority"`. An explicit
+without restarting the proxy. For example, `gpt-5.5-fast` is sent upstream as
+model `gpt-5.5` with `service_tier: "priority"`. An explicit
`codex.serviceTier` / `CCP_CODEX_SERVICE_TIER` override still takes precedence.
Reasoning effort: Claude Code's `output_config.effort` value (the one you see in
diff --git a/src/cli.ts b/src/cli.ts
index 50824b5..50e1640 100755
--- a/src/cli.ts
+++ b/src/cli.ts
@@ -36,10 +36,10 @@ async function main() {
console.log(` export ANTHROPIC_BASE_URL="http://localhost:${port}"`);
console.log(` export ANTHROPIC_AUTH_TOKEN="anything"`);
console.log(
- ` export ANTHROPIC_MODEL="kimi-for-coding[1m]" # or gpt-5.4[1m], etc.`,
+ ` export ANTHROPIC_MODEL="gpt-5.5" # or kimi-for-coding[1m]`,
);
console.log(
- ` export ANTHROPIC_SMALL_FAST_MODEL="kimi-for-coding[1m]" # background / title-gen`,
+ ` export ANTHROPIC_SMALL_FAST_MODEL="gpt-5.4-mini" # background / title-gen`,
);
console.log(` export CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC="1"`);
return;
diff --git a/src/server.ts b/src/server.ts
index 4e532cc..ce6b80c 100644
--- a/src/server.ts
+++ b/src/server.ts
@@ -191,11 +191,8 @@ function captureInboundTraffic(
ctx.traffic.writeJson("010-anthropic-request", body);
}
-// 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.
+// Claude Code may append a [1m] suffix to model names as a local compaction
+// hint. The suffix is not part of the upstream model id.
export function normalizeIncomingModel(model: string): string {
return model.replace(/\[1m\]$/i, "");
}