A dead-simple multi-agent research slash command for opencode. Designed for non-coding questions — get a synthesized, evidence-based answer from parallel AI researchers.
git clone https://github.com/brandenon/deepanswer
cd deepanswer
opencodeClone the repo or download as a folder. Start Opencode TUI in the folder or alternatively add the folder to Opencode Desktop.
Use the slash command to ask a question:
/deepanswer Is intermittent fasting supported by strong clinical evidence?
/deepanswer --deep Compare solar vs wind energy by cost and scalability
/deepanswer --experts 4 --verify What caused the 2008 financial crisis?
/deepanswer --file answer.md What are the best evidence-based sleep hygiene practices?
deepanswer will spin up two expert researchers relevant to the query by default and synthesize their outputs to answer your question.
| Flag | Effect | Default |
|---|---|---|
--deep |
4 researchers | 2 |
--experts N |
N researchers (2-10) | 2 |
--verify |
Fact-check after synthesis | off |
--file PATH |
Write answer to file | off |
--nocontext |
Skip project context | context gathered |
- Classify — maps your question to domains (science, economics, health, history, etc.)
- Select — picks the best-matching specialist agents from the agent directory
- Dispatch — launches researchers in parallel, each doing real web research
- Synthesize — converges agreements, resolves contradictions, flags gaps
- Verify (optional) — fact-checker validates key claims against sources
| Agent | Domains | Role |
|---|---|---|
domain-researcher |
adaptive | Deep research on any domain |
comparative-analyst |
comparison, science, economics, decision-making | Side-by-side evaluations |
fact-checker |
fact-checking | Called by the --verify flag when needed |
New domain specialists are auto-created when a question needs expertise not covered by existing agents.
- Add Obscura as MCP server for faster/better web fetching
- Add implementation for Claude Code and other software