one JSON API per primitive
One catalog over every subagent*.com API.
This family follows a consistent shape: each durable primitive — Contract, Task, Identity, Citation, Cache snapshot, Rubric, Outcome, Evaluation, Brand guideline, Agent Definition, Prompt Technique — ships as its own small D1-backed Cloudflare Worker with its own real GET/POST(/PATCH) /api/<resource> JSON API. That consistency is exactly what makes a unified catalog possible: subagentapi.com holds no data of its own, it just documents the pattern once and, on /health, actually calls every one of those APIs live to prove they're up.
The 37 primitives
- subagentcontracts.comThe Contract primitive — durable, typed closure over identity + persistence. definition-of-done kinds: objective_check, human_approval, human_choice, unclosable.
- subagenttasks.comThe Task primitive — generalized from the real TaskCreate/TaskUpdate/TaskList tool schema powering this session's own Progress panel. contract_id FK joins to subagentcontracts.com.
- subagentidentities.comThe Identity primitive — channel service-account vs. DM personal-account duality, grounded in Claude Tag's agent-identity docs.
- subagentcitations.comThe Citation primitive — generalized from Anthropic's real Citations API schema (char_location / page_location / content_block_location).
- subagentcache.comDurable, provenance-tagged history of real Cloudflare cache/CDN analytics for this account (manual_dashboard_paste vs. graphql_api).
- subagentrubrics.comThe Rubric primitive — typed scoring criteria used to grade agent or skill output before a merge/outcome gate.
- subagentoutcomes.comThe Outcome primitive — the typed result a manager-tier agent evaluates a coworker's work against, closing the loop a Contract opens.
- subagentevaluations.comThe Evaluation primitive — a scored pass over a Rubric against a piece of work, the record that justifies an Outcome.
- subagentbrands.comThe Brand Guideline primitive — durable design-system reference (tokens, voice, accent assignments) the rest of the family can query instead of re-deriving.
- subagentcode.comThe Agent Definition primitive — this repo's own typed Claude Code sub-agent system (AgentConfig), including the 2026-07-01 engineering-coworker upgrade to claude-fable-5.
- subagentprompts.comThe Prompt Engineering primitive — real prompting techniques and guardrails, including the Sonnet-main/Fable-advisor pairing this session actually used.
- subagentskills.comThe Agent Skill primitive — a catalog of every real SKILL.md file this repo's plugins ship.
- subagentroles.comThe Coworker Role primitive — the job-function/responsibility axis, distinct from Agent Definition (subagentcode.com).
- subagentmemories.comThe Memory Store + Dream primitives — versioned, path-addressed memories capped at 100kB/~25k tokens (memory.md), plus dreams that consolidate a store into a fresh output store without ever mutating the input.
- subagentmodels.comThe Model primitive — the real Claude models this repo's own code depends on (crates/schema/src/agent.rs's AgentModel enum, this session's own Sonnet 5, claude-mythos-5 correctly marked unused), each row citing an exact source.
- subagentrust.comThe Rust + serde primitive — real serde-deriving structs/enums grepped live from this repo's own crates/.
- subagentpython.comThe Python + Pydantic primitive — grounded in scripts/agents/schema.py, a real Pydantic v2 mirror of agent.rs added 2026-07-01 (this repo had no prior in-repo Pydantic code).
- subagenttypescript.comThe TypeScript + Zod primitive — real Zod schemas in scripts/agents/schema.ts, including a documented drift-bug fix (missing claude-fable-5) found while building this catalog.
- subagentpermissions.comThe Permission Policy primitive — always_allow/always_ask defaults per toolset and per-tool overrides, from permission-policies.md.
- subagentvaults.comThe Vault/Credential primitive — mcp_oauth/static_bearer/environment_variable credential types, from vaults.md.
- subagentsessions.comThe Session primitive — the real 4-state session lifecycle (idle/running/rescheduling/terminated), from session-operations.md.
- subagenttools.comThe Tool primitive — the real 8-tool agent_toolset_20260401, from tools.md.
- subagenthooks.comThe Webhook primitive — the 11 real session/vault webhook event types, from webhooks.md.
- subagenthandoffs.com
- subagentgraphql.comThe GraphQL types primitive — real GraphQL SDL/type patterns cataloged for the family.
- subagenthtml.comThe HTML patterns primitive — real markup/style patterns from the style-html skill.
- subagentswift.comThe Swift patterns primitive — real Swift/SwiftUI patterns from the style-swift skills.
- subagentengineering.comThe Engineering coworker role-facts primitive — real facts describing the engineering-coworker tier.
- subagentlegal.comThe Legal coworker primitive — legal-skills and legal-facts catalogs for the legal-coworker tier.
- subagentfinance.comThe Finance coworker primitive — real facts describing the finance-coworker tier.
- subagentmanagers.comThe Manager-tier primitive — the manager coworkers in the operational-plan hierarchy.
- subagentsupport.comThe Support-capabilities primitive — real capabilities this family's support surface exposes.
- subagentplugins.comThe Repo Plugins primitive — this repo's own plugins/ catalog plus external plugin tools.
- subagentsdk.comThe SDK primitive — real SDKs this repo/family depends on or documents.
- subagentservers.comThe MCP Servers primitive — real MCP servers connected across this repo's surfaces.
- subagentspec.comThe Criteria primitive — real spec/rubric criteria used elsewhere in the family.
- subagentprotocols.comThe Protocols primitive — real agent-readiness protocols (x402, AP2, WebMCP, Web Bot Auth, OAuth discovery) this family has implemented.
Explore
Browse the full endpoint catalog → · Check live health across all 37 sites → · Hit the machine-readable catalog API →