| name | agentstack-8dna |
| description | Applies AgentStack 8DNA (JSON+): structured JSON data model with built-in support for variants (e.g. A/B tests). Use when the user mentions data store, database, structured data, A/B tests, experiments, or design of AgentStack data (data/config/protected). |
AgentStack 8DNA — JSON+ (Structured JSON and Built-in Variants)
8DNA is AgentStack's data model (JSON+): structured JSON storage with built-in support for organizing data and running variants (e.g. A/B tests). Business data and config live in data (and data.config); secrets live in protected.
Data store (the "database"): Each project and each user has a JSON document. Read/write via the key-value API (GET/POST /data with keys project.data.<path>, user.data.<path>) or via project API (projects.get_project returns project including data; projects.update_project accepts data). See repo docs/architecture/DNA_KEY_VALUE_API.md.
What you can do with JSON+: Store all business data and public config in data (with optional data.config); keep secrets in protected; use dot-notation keys for consistent access; use built-in support for variants (e.g. A/B tests). APIs and MCP operate on this same structure.
When to use
- User asks where data is stored or how to use the database.
- User mentions structured data, config, or design of AgentStack
data / config / protected.
- User asks about A/B tests, experiments, or variants.
- Design or code reviews of AgentStack data and config structure.
Capabilities
- Structure (
data / protected): All business data and public config in data (e.g. data.config, data.system.config); sensitive data only in protected (keys, secrets), never logged or exposed. Dot-notation keys (e.g. config.webhooks.timeout_ms, config.permissions.read) for consistent access.
- Built-in variants: JSON+ supports organizing data and running variants (e.g. A/B tests). For advanced usage see repo docs.
Instructions
- Store business data and config in
data (and data.config); never put secrets in data — use protected and documented key paths.
- Use dot-notation keys in
data and config for consistency.
- For "where is the database?" or "how do I read/write?": use key-value API (GET/POST
/data, keys project.data.*, user.data.*) or projects.get_project / projects.update_project.
- For A/B tests or variants: JSON+ has built-in support; refer the user to repo docs for details.
Examples
- "Where is data stored?" → Project and user each have a JSON document (
project.data, user.data). Read/write via key-value API (GET/POST /data) or projects.get_project / projects.update_project. See DNA_KEY_VALUE_API in repo.
- "Store project config" → Use
project.data.config.<path> via key-value API or projects.update_project with data.
- "A/B tests or experiments?" → JSON+ supports variants (e.g. A/B); for implementation details see repo docs.
References
- DNA Key-Value API (data store) — docs/architecture/DNA_KEY_VALUE_API.md in repo: how to read/write project.data and user.data via GET/POST /data or project API.
- MCP_CAPABILITY_MATRIX — repo docs/MCP_CAPABILITY_MATRIX.md, generated from
GET /mcp/actions.
