/ai-agent-memory
AI agent memory that outlives the agent.
Most agent memory belongs to one agent. When you run several specialised agents, each rediscovers what the others already learned. Oneda gives them persistent, shared memory — retrieved by relevance, scope and permission.
Specialised agents. Shared understanding.
Agents keep their own context windows. They don't receive everything — only memories relevant to the task and allowed by their key's scope.
Product Agent
Architecture Agent
QA Agent
Oneda
Development Agent
What persistent agent memory needs
PersistenceSurvives sessions, model changes and agent rewrites.
ScopeWorkspace isolation and role-based access decide who sees what.
ProvenanceEvery memory records its source, so agents can weigh it.
HistorySuperseded decisions stay linked to what replaced them.
Connect your agents through MCP or the REST API.
One workspace, many agents, one memory.