Loading ToolBench
Seiza MCP Server and Dashboard - A multi-agent orchestration platform with DAG-based task execution, skill management, bridge server support, and web dashboard
Seiza defines 9 tools with explicit schemas and descriptions, but quality is inconsistent. Tool names follow verb_noun convention (run_, list_, get_, call_, install_) which is good. However, parameter descriptions are sparse or missing entirely. Most tools lack output schema documentation. Error handling guidance is absent. The server exposes powerful operations (run_seiza_task, call_bridge_tool, install_seiza_skill) with minimal safety guardrails or permission checks visible in the definitions. Descriptions are adequate (50-150 chars) but parameters lack the detail needed for LLM reasoning.
Directly proxies a tool execution to a downstream bridged MCP server
Returns live status, steps, and logs of active or completed tasks
Installs a skill from GitHub or local path
Returns all tools exposed by connected bridge servers
Returns list of all available agent templates in ./agents/ with their names, descriptions, models, and allowed tools
Returns list of available 9router models and current role assignments from ConfigManager
Returns list of all installed skills
Output schemas not documented. Tools return data but LLMs cannot predict field names, types, or structure. Breaks downstream tool chaining and forces agents to guess at response format.
Parameter descriptions missing or minimal. 'serverName', 'toolName', 'arguments' in call_bridge_tool lack guidance on format, constraints, or examples. 'source' in install_seiza_skill does not explain valid formats (github:owner/repo vs full URL vs local path).
No error handling guidance. Tools lack recovery hints. If run_seiza_task fails, LLM has no guidance on whether to retry, check logs, or try a different agent. No actionable error messages defined.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 51 | <=2025-11-25 | v2 |
Run a delegated seiza task
Runs a single specific agent directly
Destructive operations lack confirmation or dry-run support. install_seiza_skill and call_bridge_tool can modify state or execute arbitrary code, but no confirmation step or preview capability is documented.
No permission or scope declarations. Tools like call_bridge_tool and install_seiza_skill expose powerful capabilities without visible permission checks or scope requirements in definitions.