Agent Protocol

Continuum is not trapped inside its UI.

External AI agents connect to Continuum via the Model Context Protocol (MCP). Agents call the same state resolver, conflict detector, and evidence engine as Slack and the web application.

Claude / Agent→
Continuum MCP→
Semantic tools→
Canonical company memoryHydraDB

Wire it up

Point any agent at Continuum in one paste.

Register the Continuum MCP server with your client, or call the adapter directly. Either way, your agent gets deterministic, evidence-backed company state — the same layer Slack and the web app read.

  • ✓Same canonical state as every surface
  • ✓Deterministic — no hallucinated ownership
  • ✓Evidence + provenance on every answer
HydraDBterminal
# prerequisite: the graph is running
$ make hydradb-up
 
# start Continuum's MCP server
$ uv run continuum-mcp
✓ Continuum MCP ready — 9 tools exposed

Semantic MCP Operations:

mcp-rpc-inspector.json
JSON-RPC 2.0

// Agent Tool Request:

POST /mcp/tools/get_current_state { "entity": "account:acme", "predicate": "OWNS" }

// Continuum Deterministic State Response:

{ "value": "Priya", "valid_from": "2026-08-01", "status": "definitive" }
Company Memory

Equip your AI agents with grounded company memory.

Grounded, evidence-backed answers over one canonical memory — not fragile keyword search or a hallucinating chatbot.