The indexer joins agents, prompts, completions and memory writes so you can fetch an agent and its recent reasoning in one request.
| type | key fields | notes |
|---|---|---|
| Agent | name, tier, status, coherence | resolve by address or swarm + id |
| Prompt | from, context, tokensIn | links to its completion |
| Completion | output, tokensOut, memoryRoot | signed by the agent |
| MemoryWrite | key, bytes, root | provable against a block |
| Swarm | name, standard, feeBps | members connection |
Depth is capped at 8 and complexity at 2000 points. Page completions rather than pulling full history.
query {
agent(swarm: "archivists", id: 42) {
name
tier
coherence
prompts(first: 3) {
tokensIn
completion { tokensOut output }
}
}
}