For teams running several AI agents
Your agents talk in real time. You keep the record.
Stop copying work between agents. Connect compatible agents through an always-on listener, with LogicMem shares saved to a tamper-evident record.
Right now, you’re the go-between.
You paste one agent’s output into chat, email or a spreadsheet so the next one can use it. Those handoffs wait on you. When a handoff goes wrong, you rebuild it from chat history.
- Share once.One agent shares a decision, a finding or a task — with one agent or all of them.
- Pick up what was missed.Agents that were offline can fetch recent shares the next time they check in.
- Check it later.Shares made through LogicMem go on a tamper-evident record your team can re-check.
Sample audit record
Re-check the record anytime.
Re-check the hash-linked record for changes to recorded entries when your risk team, an auditor or a customer asks.
What setup takes
Add the LogicMem MCP connector to each agent: one URL and an API key. Keys are issued to pilot teams while self-serve sign-up is being rebuilt.
- Connect agents — Point each compatible MCP client at the LogicMem connector with the same workspace key.
- Run a handoff — One agent shares a note. Configured listeners may be notified; other agents check in.
- Re-check the record — Run a record check when someone asks to review the recorded entries.
What is protected
New memory text is encrypted with Fernet before it is stored. Fernet is AES-128-CBC with HMAC. Writes use a bearer API key, and a normal key writes only its own workspace. The audit log is SHA-256 and hash-linked.
Questions
Do my agents get a share straight away?
In our configured, always-on setup, connected listeners exchanged shares and responded to tasks in real time. Compatible agents that keep a listening stream or configured webhook can receive shares in real time; setup is pilot-assisted. Agents without a listener can fetch recent shares when they check in. We don’t publish a latency figure. A notification is not the same as the agent running a task—that depends on the agent’s platform.
What if an agent is offline?
The share stays in LogicMem. When the agent checks in, it can fetch recent shares. Check-ins return a limited recent batch, and a shorter result isn’t proof that every missed share came back. Catch-up limits are reviewed during pilot setup.
Which agents work with it?
LogicMem connects over MCP. Your agents need a compatible MCP client; other programs need integration through an adapter. Configured webhooks from the memory service are signed with HMAC-SHA256. A retry can arrive more than once, and delivery is not guaranteed. Not every HTTP client receives notifications, and we don’t claim a push test for every named product.
What goes on the record?
Writes made through LogicMem, including shares, go on the hash-linked record. Reads such as recall may be logged in the audit log rather than that chain. The record holds the notes your agents submit, not whole conversations or the actions that follow.
Does this reduce token or usage costs?
We don’t claim that. If the receiving agent acts on a share, that may use its normal usage allowance. What it removes is copying work between agents by hand.
Is this the A2A protocol?
No. LogicMem doesn’t implement the Agent2Agent (A2A) open protocol. Agents connect to LogicMem over MCP.
See it with your own agents.
Request a pilot to connect your agents, run a handoff, and re-check the record. API keys are issued to pilot teams while self-serve sign-up is rebuilt.