About this skill

Problem Solved

When local scripts, terminal workflows, or automation jobs call Coze Coding media and model capabilities, directly assembling HTTP requests usually exposes three error-prone details: personal access tokens, project context, and request/response structure. coze cli targets these workflows by wrapping the integration APIs behind coze-coding-ai into command-line capabilities, so image generation, video generation, speech synthesis, speech recognition, web search, LLM conversation, and Embedding can run from a local directory instead of scattering auth headers, project IDs, and API parameters through business code.

Core Capabilities and Mechanics

Model and media capabilities: Built on Coze Coding integration APIs, covering Seedream image generation, video generation, speech synthesis/recognition, web search, conversational LLM calls, and Embedding.
Authentication: Prefers a pat_ personal access token, supplied via COZE_API_TOKEN or .env; the SDK may also read a local runtime token.
Project context: Integration endpoints generally require a project_id; a common approach is maintaining a .coze file in the working directory with a quoted project_id, or explicitly passing a project header with -H.
Failure boundaries: Missing project context or insufficient PAT permissions can return 403; invalid or expired tokens can surface as 401 or invalid token. coze auth status only affects coze code * commands and does not determine whether this CLI is usable.

Suitability

Well suited for embedding Coze Coding media generation, search, conversation, and vectorization into local workflows, prototypes, or batch scripts. For multi-tenant services, complex permission management, or high-concurrency gateways, the surrounding system still needs rate limiting, audit logging, key rotation, and server-side authorization policies.

Screenshots

Use Cases

Generate a batch of Seedream images in a local job directory while keeping project auth via `.coze` and a PAT.
Chain web search and LLM conversation in the terminal to validate RAG prompts and response fields.
Prepare speech synthesis audio and recognition text samples, then tune speech API parameters.
Batch-generate embeddings for source documents to compare retrieval quality and vector dimensions locally.

Best For

Backend engineers who need to call Coze image, video, and speech APIs from local scripts.
Full-stack engineers debugging LLM, web search, and Embedding calls in the terminal.
Automation engineers maintaining Coze Coding auth and project context for workflows.
Product engineers validating local prototypes that need quick access to media generation APIs.