Architecture · LikeC4
databot-agent
Tool-agnostic client layer: one CLI plus agent assets that route data questions to the data-query service and shape the answer
The whole picture
System landscape
The system in context, then opened up into its containers.
databot-agent — system landscape
index
databot-agent — client surfaces (CLI, Claude assets, MCP plugin)
databotSystemTool-agnostic client layer: one CLI plus agent assets that route data questions to the data-query service and shape the answer
Risks & open questions
risksInside each box
Containers & components
Each part decomposed into the components that implement it. Every box links to its source in the interactive explorer.
databot CLI — internals (auth, transport, parsing)
cliContainerSingle-file Python 3 script (stdlib only). Resolves auth, POSTs the question, and shapes the JSON envelope. Every other surface ultimately calls this
Claude Code assets — subagent + three skills
claudeAssetsContainerA subagent and three skills, all markdown + YAML frontmatter, loaded on demand by question shape. Encodes the routing, validation, and redaction discipline around the raw CLI
Cowork MCP plugin — stdio server + tool + skill
coworkPluginContainerA stdio MCP server (stdlib Python) that wraps the CLI as a single `databot_query` tool, plus a bundled skill. Lets an MCP host call the backend with no per-session token pasting
How it runs
Walkthrough flows
Dynamic views — the narrative spine of the system, step by step.
Walkthrough — a direct CLI question end-to-end
cliFlowThe simplest path: an analyst asks a question in a terminal. The CLI resolves identity locally, POSTs once, and prints the JSON envelope (answer + audit trail).
Walkthrough — the subagent research workflow
researcherFlowThe opinionated path: a host agent dispatches one self-contained question. The subagent pre-routes, probes, checks provenance, validates for known quirks, redacts PII, and returns a terse cited answer — keeping the large raw audit trail out of the orchestrator.
Walkthrough — the MCP tool path
mcpFlowThe embedded path: an MCP host calls databot_query. The stdio server shells out to the same CLI, summarizes the audit trail by default, and returns the answer inline — no per-session token pasting.
Where it runs
Deployment
What runs where, and the process & data boundaries between the pieces.
Deployment — where each piece runs
deploymentExplore it live
The figures above are static exports. The interactive explorer lets you pan, zoom, follow relationships, and jump from any box to the source.