@cloudflare/codemode@0.5.4
cloudflare/agents@cloudflare/codemode@0.5.4Oct 13, 2026by github-actions[bot]
Upgrading?
AI Summary
Enhances `ToolSetConnector` to properly forward output schemas, ensuring `codemode.describe()` displays accurate return types instead of `unknown`. It also enforces the requirement for `.default()` fields and adds support for `.pipe()` types.
Key Highlights
- `ToolSetConnector` now forwards `outputSchema`, fixing `codemode.describe()` to show actual output types.
- Async schemas (`jsonSchema(promise)`) are now awaited instead of being dropped.
- Output types now describe the return value of a tool rather than the input.
- `.default()` field is now required, and `.pipe()` support is added for piped types.
- .transform() types are represented as `unknown` due to JSON Schema limitations.
New Features
- Improved output schema display in connector descriptors.
- Awaiting of asynchronous JSON schemas.
- Added `.pipe()` type support.
- Enforced `.default()` field requirement.
Full Release Notes
### Patch Changes
- [#2472](https://github.com/cloudflare/agents/pull/2472) [`7945da9`](https://github.com/cloudflare/agents/commit/7945da9c6e142119e296984fc9caadec41e182a5) Thanks [@mattzcarey](https://github.com/mattzcarey)! - `ToolSetConnector` now forwards each AI SDK tool's `outputSchema` into its connector descriptors, so durable `codemode.describe()` shows the output type instead of `unknown`, as `McpConnector` already does. Input and output schemas that convert asynchronously (`jsonSchema(promise)`) are now awaited instead of being dropped.
Output types, in both `generateTypes` and connector descriptors, now describe the value a tool returns rather than the input side of its schema. A `.default()` field is now required and a `.pipe()` field has the piped type. A `.transform()`, which JSON Schema can't express, shows as `unknown` instead of the pre-transform type.
Related Repositories
The AI that really does things. Any OS. Any Platform. The lobster way. 🦞
Interactive roadmaps, guides and other educational content to help developers grow in their careers.
Reverse engineer anything with agents, from app behavior down to native binaries.
The open source coding agent.
Fair-code workflow automation platform with native AI capabilities. Combine visual building with custom code, self-host or cloud, 400+ integrations.