Wiki index

The content catalog for the knowledge wiki. The agent updates this on every ingest. Browse at /wiki/<slug>; see log.md for the timeline.

Overview

• overview - synthesis of the agent tooling stack

Concepts

• context management - keeping a codebase in mind without drowning the model
• BYOK and model choice - bring-your-own-key and model freedom
• MCP - the Model Context Protocol as a shared tool interface
• self-hosting - running tooling on your own infrastructure
• git-native workflows - git as audit log and rollback
• evaluation and observability - tracing is not evaluation
• comprehension debt and skill atrophy - the cognitive cost of agentic coding
• agent leverage and loops - success criteria over instructions
• context as the bottleneck - shared context, not code, is the limit
• code review with agents - self-improving and multi-lens review
• agent experience - designing codebases agents work well in
• specs and planning - plan before you generate
• structured input and output - structure in, structure out

Entities

• Claude Code - Anthropic’s terminal coding agent
• Aider - git-native terminal coding agent
• OpenCode - provider-agnostic, local-first coding agent
• Langfuse - open-source LLM observability