SQUAD MEMORY

Persistent shared memory for your AI agents.

Save a decision with Claude. Pick it up with Codex. Squad Memory keeps project decisions, discoveries and next steps available across compatible agents and sessions on your TermSquad Computer.

  • Across compatible agents
  • Across sessions
  • On your computer
One project. Shared knowledge.Example workflow
Claude CodeSession 01 · Working on authentication
01

Saves a project decision

Squad Memoryyour-app
PROJECT DECISION

Reuse the existing
auth boundary.

Keep session handling consistent
with the rest of the application.
architectureauthentication

Retrieves the saved decision

CodexSession 02 · Continuing the same project
02
The conversation changes. The decision stays.

What is Squad Memory?

Squad Memory is TermSquad’s persistent shared memory for AI agents, sessions and harnesses.

What your agents can remember

The decisions. The discoveries.
The way forward.

Your code tells part of the story. Save the useful knowledge behind it, so another agent can pick it up.

Decisions

Keep the reasoning.

Save what you decided and why, so the next agent can work with the same constraints.

Example memory
Reuse the existing auth boundary.
Keep session handling consistent across the app.architecture
Discoveries

Carry the lesson forward.

Keep useful findings and the root causes behind fixes available for the next task.

Example memory
The timeout came from a stale connection.
Reconnect before retrying the request.bug fix
Handoffs

Leave a clear next step.

Record what was accomplished, relevant files and what still needs attention.

Example memory
Auth changes are ready for review.
Next: test the expired-session flow.session summary

How Squad Memory works

Save it once.
Build on it next time.

Compatible agents use the memory integration to save and retrieve knowledge for the project they are working on.

  1. 01

    Save what matters.

    As work progresses, your agent saves useful decisions, discoveries and the reasons behind them to the project’s shared memory.

  2. 02

    Find the relevant context.

    In a later session, a compatible agent retrieves recent context and searches for the project knowledge relevant to its task.

  3. 03

    Continue with the knowledge.

    The agent uses the saved knowledge to continue the work, updates it as the project changes, and leaves a summary with next steps.

A new session, with contextEXAMPLE
You ask

“Continue the authentication work.”

Your agent retrieves saved project context

Decision
Reuse the existing auth boundary.
Why
Keep session handling consistent.
Next step
Test the expired-session flow.
A handoff starts from what was saved.
01

Search by topic

Find saved decisions and discoveries with full-text search, then read the complete observation.

02

Recover after compaction

Retrieve saved project context when a long conversation is compressed or a fresh session begins.

03

Keep knowledge current

Update evolving topics through revisions, with deduplication to reduce repeated memories.

Full-text and contextual search, with agent-assisted semantic relations where configured.

Your computer. Your project knowledge.

Shared by your agents.
Stored on your computer.

Squad Memory lives on your TermSquad Computer. Compatible agents on that machine use the same store, with context scoped to the relevant project.

Memory content is not sent to the TermSquad control plane. When your agent retrieves it, that context can be processed by your chosen AI provider.

Explore security and data boundaries

Two kinds of continuity.

They work together, each keeping something different.

Persistent Sessions

Keep the work running.

Disconnect and return to the same running workflow.

Squad Memory

Keep the knowledge available.

Start a different session and retrieve saved project context.

Frequently asked questions

A few things
to keep in mind.

What gets saved, how agents retrieve it, and where your knowledge lives.

Explore the documentation
What is AI agent memory?

AI agent memory is saved knowledge that an agent can retrieve beyond one conversation or context window. It can include project decisions, discoveries, session summaries and next steps.

Can Claude and Codex share the same memory?

Yes. Compatible agents such as Claude Code and Codex can save and retrieve project knowledge through the same Squad Memory store on your TermSquad Computer. The agent needs the configured memory integration and must use it to save or retrieve context.

Does Squad Memory save every conversation automatically?

Squad Memory is built around useful knowledge that agents explicitly save and retrieve: decisions, discoveries and summaries. It does not guarantee a complete transcript of every conversation or automatic recall of everything an agent has seen.

What happens after a new session or context compaction?

Saved knowledge remains available independently of the conversation. A compatible agent can retrieve recent context, search for relevant memories and read the saved details. Recovery depends on what was saved and on the agent using the memory integration.

How does memory search work?

Squad Memory supports full-text search and retrieval of recent project context. Agents can search for a topic and read the full saved observation. Agent-assisted semantic relations are available where configured; they are separate from full-text retrieval.

Where is Squad Memory stored?

Squad Memory lives on your TermSquad Computer. Its memory content is not sent to the TermSquad control plane. When an agent retrieves a memory, it can become part of the context processed by your chosen AI provider.

Is Squad Memory the same as persistent sessions?

Persistent Sessions preserve the work. Squad Memory preserves the knowledge. A persistent session keeps a running workflow available after you disconnect; memory keeps saved project knowledge available when an agent or conversation changes.

Is memory shared across projects or computers?

Compatible agents on the same TermSquad Computer use one shared store, with memories scoped to the relevant project. Agents should retrieve context for the project they are working on. This does not imply automatic memory synchronization across separate TermSquad Computers.

The next agent starts ahead

New session.
Keep the understanding.

Give your agents a place to build on what came before.

AI subscriptions and usage are not included.