Skip to main content

Documentation Index

Fetch the complete documentation index at: https://ahvn.top/llms.txt

Use this file to discover all available pages before exploring further.

HeavenBase has two MCP paths: workspace tools for data, and Capsule Toolkits for Python functions.

​
1. Choose a surface

Use Workspace MCP when an agent needs to inspect entities, define schemas, write rows, run JSON queries, explain routing, or use the memory profile. Use MCP in HeavenBase when you need to package reviewed Python callables as a Toolkit, persist them in the registry, and serve them through FastMCP.

​
2. Generate client config

Workspace servers and Toolkit servers both expose to_mcp_json(...): 
import heavenbase as hb

workspace = hb.HeavenBase("notes", preset="debug")
print(workspace.to_mcp_json(name="notes", transport="http", host="127.0.0.1", port=7011))

toolkit = hb.Toolkit.load(name="math-tools", namespace="demo", version="1")
print(toolkit.to_mcp_json(transport="stdio"))
Use HTTP when a long-running process owns the server. Use stdio when the MCP client should launch HeavenBase in the active Python environment.

​
3. Keep tool surfaces small

For agent-facing workspace access, prefer profile="agent", profile="memory", or profile="memstate" before exposing profile="full": 
agent_tools = workspace.to_mcp(name="notes-agent", profile="agent")
memory_tools = workspace.to_mcp(name="notes-memory", profile="memory")
memstate_tools = workspace.to_mcp(name="notes-memstate", profile="memstate")
Use profile="memory" for a small note surface. Use profile="memstate" when the agent needs project keypaths, current-state browsing, version history, conflict visibility, and soft deletes. The full workspace profile includes delete and bulk operations. Keep it for trusted application code, tests, and administrative flows.

​
Further Exploration

Related resources: