AgentScope Studio
https://github.com/agentscope-ai/agentscope-studio- Category
- AI Agents
- Rank
- No. 980Tools index
Previous survey · No. 1057 ·
- Listed in
- #6 Observe an LLM application
- Pricing
- Open Source
- Type
- TOOL
- Builder
- agentscope-ai
- GitHub
- 647 stars
- Latest release
- v1.0.9
- Date
About
Development-oriented visualization toolkit for AgentScope. Trace, debug, and monitor multi-agent LLM systems with an OpenTelemetry-aware UI.
What it does
AgentScope Studio runs beside an AgentScope application and turns its activity into an interactive workspace. It groups projects into runs, shows live agent conversations, records model calls and token use, and provides statistical evaluation views. Applications connect by supplying the Studio address, while traces arrive through OpenTelemetry endpoints. A bundled assistant, Friday, adds an AgentScope-focused playground and development help.
Why it's ranked here
A strong choice for developers already committed to AgentScope. It combines live interaction, persisted run data, trace inspection, evaluation, and an assistant in one local package. OpenTelemetry ingestion over HTTP and gRPC makes the tracing layer concrete, while npm and Docker installation options reduce setup friction. Its value narrows sharply outside the AgentScope ecosystem.
What's good
The tool covers the full debugging loop: observe conversations, inspect agent and model spans, compare run statistics, then use Friday for follow-up development work. It accepts OpenTelemetry data through both HTTP and gRPC, stores structured records locally, and uses real-time sockets for active sessions. Friday supports five model provider families and keeps file writing behind an explicit permission flag.
Tradeoffs
This is an AgentScope companion, not a framework-neutral agent console. The standard install requires Node.js 20 or newer and npm 10 or newer. Friday adds Python and AgentScope dependencies, requires provider and model configuration, and can execute Python or shell commands. Although file writing is optional, enabling it expands the assistant's local capabilities and deserves careful handling.
How to use it well
Use it while building or debugging AgentScope applications on a developer machine. Connect the application to the local server, inspect each project run, follow conversations live, and use traces to investigate model calls, tokens, and agent activity. Bring Friday in for AgentScope questions or controlled local task work. Choose another product if you need framework-neutral observability or a hosted operations service.
Technical notes+
The root package.json defines an npm workspace package, an as_studio binary, Node.js and npm engine requirements, and builds client, server, and Python assets into dist. packages/server/src/index.ts starts Express with tRPC at /trpc, raw OpenTelemetry HTTP ingestion under /v1, a separate gRPC receiver, Socket.IO, static production assets, automatic port selection, and graceful shutdown. packages/server/src/database.ts initializes TypeORM entities, migrations, and synchronization, with better-sqlite3 listed as a dependency. packages/client/src/App.tsx composes React Query, tRPC, sockets, internationalization, notifications, routing, and Ant Design. packages/app/friday/main.py creates a ReActAgent with local execution and inspection tools, optional file writing, session persistence, hooks, and real-time steering. packages/app/friday/model.py maps DashScope, OpenAI, Ollama, Gemini, and Anthropic to AgentScope chat models and formatters.
Observed
- License
- Apache License 2.0
- Languages
- TypeScript and Python
- Packaging
- npm workspace package published as @agentscope/studio
- Install surface
- Global or local npm installation, source installation, and optional Docker deployment
- Interfaces
- CLI, browser UI, tRPC HTTP endpoints, Socket.IO, and OpenTelemetry HTTP and gRPC ingestion
- Runtime requirements
- Node.js 20 or newer and npm 10 or newer
- Operating model
- Local server with persisted database records and production static web assets
Read from README.md, package.json, packages/client/vite.config.ts, packages/client/eslint.config.js, packages/app/friday/main.py, packages/client/src/App.tsx, packages/app/friday/args.py, packages/app/friday/hook.py, packages/client/src/main.tsx, packages/server/src/index.ts, packages/shared/src/index.ts, packages/app/friday/model.py, packages/client/src/styles.ts, packages/server/src/database.ts, packages/client/src/vite-env.d.ts.
What it can do
Trace multi-agent LLM system execution
Multi-agent LLM system with OpenTelemetry instrumentation → Execution traces showing agent interactions and workflows
Debug multi-agent conversations
Agent interaction logs and conversation flows → Visual debugging interface with conversation breakdowns
Monitor agent system performance
Real-time agent system metrics and telemetry data → Performance dashboards and monitoring visualizations
Visualize agent communication patterns
Agent-to-agent communication data → Visual network diagrams of agent interactions
Analyze LLM system bottlenecks
System performance metrics and execution traces → Bottleneck analysis reports and optimization suggestions
Display OpenTelemetry spans and metrics
OpenTelemetry telemetry data from AgentScope applications → Structured visualization of spans, metrics, and traces
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.