Developer builds own coding agent harness, fights short context limits of free LLM tiers
danja · reddit · 2026-09-15
The author's old CPU can't run Claude Code, OpenCode or Codex, so they built an open-source harness called "peasant" (github.com/danja/peasant), wiring it to free-tier services like Groq and Mistral. It works, but the short context windows of these services frequently cause the agent to get stuck; they're asking for engineering strategies to handle short-context scenarios.
More from coding & agent
- Gemma via Exa MCP insists it's May 2024, returns stale data — iaderia · 2026-09-15
- Anthropic engineer behind Building Effective Agents explains agents in 14 minutes — HeyAmit_ · 2026-09-15
- Four dev boards hooked to the internet: test AI-written firmware on real silicon via HTTPS — SelfishlyWandering · 2026-09-15
- 'Mom, shhh, I'm talking to Devin AI': the meme every dev relates to — marvinvonhagen · 2026-09-15
- Social Quack launches: buy-once Mac scheduler with MCP hookup to Claude and ChatGPT — ThePeterMick · 2026-09-15
- Why enterprise AI pilots build a "tool museum" instead of capability — and the agentic-OS fix — Hamza_StrategizeLabs · 2026-09-15