THE SYSTEM OF RECORD FOR AI CONTEXT
Conxt gives people and agents one trusted context layer. Every decision, rule, and preference is scoped, attributed, and available across the tools your team already uses — without losing who wrote it or why.
A human recorded it. An agent used it. The provenance stayed attached.
One substrate
People and agents read and write the same governed record.
Every write attributed
Human, agent, source, session, and time remain attached.
Nothing silently replaced
Contradictions are quarantined or sent for human resolution.
Works across the tools your team already uses
The problem
A decision made in Claude never reaches the agent running in production. Shared state changes, but no one can reconstruct which actor wrote it, what another agent relied on, or where a bad output began. Transport protocols connect agents; they do not govern what those agents know.
What governed memory gives you
Follow one decision — from the moment a person records it, to an agent acting on it, to the day it's challenged.
Use pdf-lib over pdfkit.
Priya · ClaudeRecorded once by a person — available to every approved tool and production agent.
An agent reads it before acting.
Build agent · Cursor · run-0706Actor, source, scope, and time move with it — even across tools.
Contradictions are flagged; a human chooses what wins. Conxt never guesses.
Shared, yet private
Everyone inherits the company's standards. Teams share their own context. And what's personal stays personal — until you choose to share it up. The right people see the right memory, and nothing else.
Company standards and decisions.
A group’s shared context, across all their tools.
Private by default — share it up when it should travel.
Each layer sees everything above it — never below.
Over MCP
Conxt is a Model Context Protocol server. Any MCP client — Claude Desktop, Cursor, or your own agents — connects with a key and reads only what policy allows. No SDK lock-in, no bypass.
get_contextAgents pull scoped, policy-filtered context before they act.
write_memoryNew memories are attributed to the agent and session — risky writes are quarantined for human review.
recall_as_ofReconstruct exactly what an agent knew at any point in time.
list_conflictsSurface contradictions in a scope before they cascade.
Pricing
Team
$25/ seat / month
For a team standardizing on one shared memory.
Enterprise
Design partnerCustom
Org-wide governance, built with you as an early partner.
Just you? The free plan covers individual memory.
A 20-minute walkthrough on your team's real decisions. Start with one team, and expand when it earns it.