Summary
Contribute a DocsClaw-based agent example to the
kagenti/agent-examples
repository, demonstrating a Go-based alternative to the Python agents.
Motivation
All current kagenti agent examples use Python with various
frameworks (LangGraph, AG2, CrewAI, Marvin). DocsClaw can
replace 6 of the 10 agents with configuration alone — different
ConfigMaps, same 10 MiB binary. This demonstrates:
- Single runtime, multiple personalities (vs. framework per agent)
- 10 MiB per pod vs. 200-500 MiB for Python agents
- No framework choice paralysis — config-driven behavior
Approach
Start by reproducing the Generic Agent — it's the most
flexible example and shows dynamic tool loading. A DocsClaw
ConfigMap with the same system prompt and MCP tool connections
would produce equivalent behavior.
Depends on
Summary
Contribute a DocsClaw-based agent example to the
kagenti/agent-examples
repository, demonstrating a Go-based alternative to the Python agents.
Motivation
All current kagenti agent examples use Python with various
frameworks (LangGraph, AG2, CrewAI, Marvin). DocsClaw can
replace 6 of the 10 agents with configuration alone — different
ConfigMaps, same 10 MiB binary. This demonstrates:
Approach
Start by reproducing the Generic Agent — it's the most
flexible example and shows dynamic tool loading. A DocsClaw
ConfigMap with the same system prompt and MCP tool connections
would produce equivalent behavior.
Depends on
MCP tool servers the Python agents use