PuppyIP Resource Center
Developer Tool Updates 5 min read Published 2026-11-12

Codex 0.167.2: Vertex incremental tools and how to opt out

Codex CLI 0.167.2 enables incremental tool declarations by default on Google Cloud Vertex AI. Developers using Vertex should check their actual version and gcp provider before deciding whether to keep the behavior; opting out uses the separate incremental_tools_legacy_gcp flag.

Codex CLI 0.167.2 Vertex AI Incremental tools Configuration

Service eligibility and regional restrictions

PuppyIP serves only compliant overseas businesses and their authorized personnel. Proxy services are not available in mainland China. The service may only be used for lawful business activities outside mainland China. Use of this service within mainland China is prohibited.

Hosting a proxy IP or server overseas does not change these restrictions. The service must not be provided to end users in mainland China through relaying, forwarding, sharing or resale. Before use, read the Terms of Service.

Key Takeaways

  • Version 0.167.2 was published as a stable release on November 11, 2026, at 22:44:46 UTC; this default change concerns Google Cloud Vertex AI.
  • On the regular Responses route, the exact gcp provider ID selects incremental_tools_legacy_gcp, which defaults to true and is marked Stable.
  • To opt out, set incremental_tools_legacy_gcp = false under [features] in config.toml; Bedrock and Responses Lite use different flags.
  • A tool declaration change does not mean a tool has executed or establish new model access, billing terms, a usage reset, or progress on Tibo's promise.

Which requests change in 0.167.2?

OpenAI published Codex CLI 0.167.2 on November 11, 2026, at 22:44:46 UTC. Its release notes say Google Cloud Vertex AI now receives incremental tool updates by default. Tool declarations describe the available tools and their parameters to the model; updating those declarations is separate from invoking a tool or running a command.

Developers already using Codex through Vertex should review this patch. A stable release does not mean the CLI on your machine has upgraded or that an account has gained access to a model or region. This article did not make real Vertex requests or measure performance or billing changes.

Check the gcp identifier and actual request route

The provider definitions in the 0.167.2 tag assign Google Cloud Vertex AI the ID gcp. On the regular Responses route, the effective model_provider_id must match that exact value to select the IncrementalToolsLegacyGcp feature. An endpoint's compatible interface or a name containing Vertex does not establish that match.

Responses Lite is a separate case. The code checks use_responses_lite first and selects incremental_tools for that route, rather than the Vertex-specific flag. Record the actual client version, effective provider ID, request route, and applied configuration when investigating; finding gcp in one configuration file is insufficient.

How do you disable the Vertex default?

The official release notes give this opt-out: set incremental_tools_legacy_gcp = false under [features] in config.toml. If [features] already exists, change the relevant key in that section instead of adding a duplicate section. Keep a copy of the original configuration so you can restore it.

For example, a team comparing the previous tool-catalog layout with incremental declarations could explicitly set false in an authorized isolated configuration, then inspect requests after that configuration has taken effect. This is a configuration example, not an experiment performed for this article. To restore the 0.167.2 default, remove the explicit false override or set it to true, then check the effective value.

Why is the Bedrock flag different?

Version 0.167.2 still selects incremental_tools_legacy for amazon-bedrock and amazon-bedrock-runtime, with a default of false. It is independent of incremental_tools_legacy_gcp, which defaults to true. Enabling the Bedrock flag alone does not control the Vertex behavior.

The feature table in the same tag also marks the Bedrock legacy flag Stable. Its UnderDevelopment status in 0.167.1 describes that older version and should not be carried forward as the current 0.167.2 status. Stability and the default are separate: the Bedrock flag is marked stable while remaining disabled by default.

How can you check whether the change applies?

First identify the binary version that is actually running and the provider branch selected by its configuration. In an authorized test environment, compare requests before and after a tool-catalog change: look for additional_tools declarations in history and check that the tools visible to the model match the available catalog. Sanitize diagnostic material before sharing it; do not publish tokens, private prompts, or full task contents.

The upstream comparison includes GCP default-on and explicit opt-out test scenarios and checks the independent Bedrock flag. Those describe the repository's validation scope, not acceptance of your environment. A tool appearing in a catalog, a successful request, and a successful tool execution must also be assessed separately.

Which benefits and entitlements remain unconfirmed?

The release notes and tagged code provide no measurements of latency, token usage, cache hits, or cost savings. They do not announce new account eligibility, regional access, or free quota. Check model access, support, and billing against the actual Vertex service terms; changing a network exit does not replace those conditions.

This is an independent Codex CLI release, not evidence that Tibo has announced delivery of a reset. It also does not explain every Workspace Agent startup failure. For a service problem, check the relevant official status and the original task's outcome separately before retrying an external action.

Sources

Frequently Asked Questions

Will a custom provider alias enable this by default?

An alias alone does not establish that. The regular Responses branch in 0.167.2 checks whether the effective model_provider_id is exactly gcp, while Responses Lite selects incremental_tools separately. Check the effective identifier and route before deciding which flag applies.

Does disabling this flag revoke model access or reset usage?

This flag controls the layout of tool declarations. The official release does not describe it as an account-access or usage-reset setting. It cannot be used to request model access, change billing eligibility, or establish that a reset has arrived.