MCP resources
Retrieve active Decisions and their history through stable resource URIs.
Give agents context they can trust.
Authenticated agents can search decisions, submit sources, and create candidates without bypassing human review or audit.
Play Decision Log dashboard product tour
Decide exactly what agents can do
Scope each capability, attribute every action, and require review before an agent changes active meaning.
Permission-scoped tools
Let each registered client use only its permitted searches and mutations.
Candidate-first output
Hold agent proposals as DecisionCandidates until a reviewer approves them.
Actor attribution
Distinguish User, Agent, API key, CLI, and MCP client activity.
Source submission
Keep extracted suggestions linked to the evidence that produced them.
Audited actions
Record sensitive MCP and agent activity in the append-only audit stream.
Review before meaning changes
Extraction and agent workflows create DecisionCandidates. They never edit active Decisions silently.
- Approve or reject with explicit permissions
- Convert a candidate into a supersession
- Retain source and Agent provenance
Connect agents without surrendering control
Register an MCP client and make the decision record available through scoped tools.
