@cloudflare/think@0.5.0

cloudflare/agents@cloudflare/think@0.5.0Apr 30, 2026by github-actions[bot]

AI Summary

This minor release (0.5.0) of @cloudflare/think introduces agent tool orchestration for running Think and AIChatAgent sub-agents as retained, streaming tools from a parent agent, along with reasoning controls and telemetry forwarding capabilities.

Key Highlights

Agent tool orchestration enables running Think and AIChatAgent sub-agents as retained, streaming tools
New APIs include `runAgentTool`, `agentTool`, parent-side run replay and cleanup
Added `sendReasoning` controls with instance-wide defaults and per-turn overrides via `beforeTurn`
Forwarded `TurnConfig.experimental_telemetry` to Think's internal AI SDK `streamText()` call
Think and AIChatAgent child adapter support with headless React/client event state helpers

New Features

Agent tool orchestration for sub-agents (retained, streaming tools from parent agent)
`runAgentTool` and `agentTool` APIs for parent agent orchestration
Parent-side run replay and cleanup functionality
Think and AIChatAgent child adapter support
Headless React/client event state helpers
`sendReasoning` controls with configurable instance defaults and per-turn overrides
Experimental telemetry forwarding to AI SDK's `streamText()` for per-turn LLM observability

Full Release Notes

### Minor Changes

-   [#1421](https://github.com/cloudflare/agents/pull/1421) [`1b65ff5`](https://github.com/cloudflare/agents/commit/1b65ff5550f904e2a59bd6015703f82b02f85e4f) Thanks [@threepointone](https://github.com/threepointone)! - Add agent tool orchestration for running Think and AIChatAgent sub-agents as
    retained, streaming tools from a parent agent. The new surface includes
    `runAgentTool`, `agentTool`, parent-side run replay and cleanup, Think and
    AIChatAgent child adapter support, and headless React/client event state
    helpers.

### Patch Changes

-   [#1424](https://github.com/cloudflare/agents/pull/1424) [`58ca2fc`](https://github.com/cloudflare/agents/commit/58ca2fc1edda0f8a91ddce853014f8a7c8662f64) Thanks [@threepointone](https://github.com/threepointone)! - Add `sendReasoning` controls to Think. Subclasses can set an instance-wide default, and `beforeTurn` can return a per-turn override to include or suppress reasoning chunks in UI message streams.

-   [#1423](https://github.com/cloudflare/agents/pull/1423) [`0ed42a9`](https://github.com/cloudflare/agents/commit/0ed42a908ed28181d12dfaa9c97e182e831d0218) Thanks [@threepointone](https://github.com/threepointone)! - Forward `TurnConfig.experimental_telemetry` to Think's internal AI SDK
    `streamText()` call so applications can configure per-turn LLM observability.