kuluru vineeth

A book you read by building

Inside a Coding Agent

Build the harness end to end, from the tool loop to an agent that ships real code.

Coding agents are treated as magic. This book takes one apart. You build a real agent one capability at a time: the tool loop, event-sourced sessions, a fail-closed sandbox, context assembly, subagents, until it takes an engineering task, edits code across a repo, runs commands, and verifies its own work. It runs on the engine you built in the first book.

Every listing on these pages is pulled from a real, working repository at a pinned commit, and every command's output was captured from actually running it. Nothing here is typed from memory.

Start reading →Free to read · 3 chapters live

Part I · The Loop

Part II · The Durable Session

  • 05Event-Sourced Sessions: Append, Resume, Forksoon
  • 06Deterministic, Byte-Exact Replaysoon
  • 07The Provider Seamsoon

Part III · The Trust Boundary

  • 08The Fail-Closed Sandboxsoon
  • 09Policy-Gated Approvalssoon
  • 10Framing Tool Output as Untrusted Datasoon

Part IV · The Context

  • 11Context Assembly and Progressive Disclosuresoon
  • 12History Compactionsoon
  • 13Durable Memory and Output Spillsoon
  • 14The Ranked Repo Mapsoon

Part V · Scaling Out

  • 15Subagents and Fork-Seeded Contextsoon
  • 16Prompt Caching, Verified End to Endsoon
  • 17MCP: Tools over stdio JSON-RPCsoon
  • 18Workflows and the Ralph Loopsoon
  • 19The Plugin Seamsoon

Part VI · Trust, Interface, and Proof

  • 20The Hardening Passsoon
  • 21The Session Viewer: Timeline and Transcriptsoon
  • 22The Session Viewer: Analytics and Live Chatsoon
  • 23The Eval Harness: Forge vs a Hosted Modelsoon