opencode

Use agent-handover with OpenCode

This example wires agent-handover into OpenCode so a session resumes exactly where the previous one stopped — even if the previous handover crashed halfway.

OpenCode reads AGENTS.md as its rules/bootstrap file (the same mechanism as Codex CLI), so this adapter shares its block and installer with the Codex adapter — only the example path and engine tag differ.

1. Install

pip install agent-handover

2. Add the bootstrap block to your repo's AGENTS.md

OpenCode reads AGENTS.md at the repo root (and a global ~/.config/opencode/AGENTS.md). Install the handover block (idempotent — safe to re-run, preserves your other rules):

from agent_handover.adapters.opencode import install_agents_md
install_agents_md("AGENTS.md")

This inserts the block shown in AGENTS.md: at session start OpenCode runs agent-handover check; on exit 1 it finishes the interrupted handover before doing new work.

3. Write a handover at the end of each session

python examples/opencode/end_of_session.py "what happened this session"
# local only (no git push):
AH_PUSH=0 python examples/opencode/end_of_session.py "wip: parser refactor"

Under the hood (end_of_session.py) this calls build_opencode_handover_engine(...), which writes a Layer-1 session note and the Layer-2 current-state.md, then publishes memory/ via git — each step checkpointed, with the PAUSE file honored as a kill-switch.