changelogAnthropic

Cline CLI v3.0.54 Fixes Claude Code Provider Tool-Bridging Bug and Token Telemetry Overcounting

TL;DR

Cline's CLI v3.0.54 release fixes a major bug that made the Claude Code provider unusable for agentic work, along with tool-call JSON truncation handling and token telemetry that was overcounting cache-heavy sessions by roughly 5x. The update also changes how Managed Hub daemons handle version conflicts between multiple Cline installs.

2 min read
0

Cline Ships CLI v3.0.54 With Claude Code Provider Fix and Telemetry Correction

Cline released CLI v3.0.54 on August 13, patching a bug that made its Claude Code provider unusable for agentic work, along with several tool-call handling and telemetry fixes pulled from SDK v0.0.74.

What broke and what's fixed

The Claude Code provider previously received tool definitions it could not bridge, breaking agentic workflows. According to the release notes, the provider now runs its own native tools instead. The session is also anchored to the workspace directory rather than inheriting the host's current working directory, and Cline now loads ~/.claude plus project-level settings so existing permission rules apply. File edits inside the workspace are auto-approved, while command execution remains gated by the user's own Claude settings.

Several tool-call parsing bugs were also fixed:

  • Truncated tool-call JSON with an unterminated string was previously "repaired" with an invented terminator, silently producing wrong arguments. It is now rejected outright.
  • Strict providers were rejecting turns with a "user message must have content" error when a message's content held only empty text parts.
  • A mid-turn crash occurred on streamed tool calls with non-zero or non-contiguous indexes, triggered through LiteLLM's Anthropic passthrough.

Token telemetry correction

Cline says per-task token totals were inflated by roughly 5x on cache-heavy sessions because the system re-counted the entire cached conversation on every request. Telemetry now reports disjoint buckets for uncached input, cache-read, and cache-write tokens instead.

Hub daemon changes

Managed Hub daemons now upgrade directionally: when another Cline install ships a newer Hub build, the CLI attaches to that newer daemon and prompts the user to update and restart, rather than two installs repeatedly retiring each other's daemons. Yolo and sandbox sessions, which don't attach to the shared Hub, are unaffected by this prompt. Separately, a bug causing the Hub daemon to log an unhandled "hub server close failed" error and exit non-zero when a client was still connected at shutdown has been fixed. Upgrading the CLI now also retires any running Hub daemon and respawns it on the new code, instead of leaving the upgraded CLI talking to a daemon still running the previous release.

Full changelog: cli-v3.0.53...cli-v3.0.54

What this means

This is a maintenance release, not a new model or feature launch, but it addresses a functionally significant defect: a provider that was reportedly unusable for its core purpose. Teams running Cline's CLI with the Claude Code provider in agentic pipelines were likely hitting broken tool execution and inflated token cost reporting — the latter potentially skewing budget and usage monitoring by roughly 5x on cache-heavy workloads. Organizations relying on Cline for automated coding tasks should update promptly, particularly if they've been troubleshooting unexplained token cost spikes or Claude Code provider failures in multi-instance Hub setups.

Comments

Loading...