Capture useful context.
Supported workflows can capture goals, decisions, project state, and next actions without turning memory into another task.
Smart shared memory · From $290/mo
Every agent. Every vendor. One memory.
Your agents from different vendors and different models share one centralized memory – encrypted end-to-end, private by design, and built to just work. What one agent learns, every agent remembers.
Two-minute install – one command in your terminal, or a macOS installer.
Why it exists
Agent context should be available whenever work needs it.
Gitmas captures useful context from supported workflows and makes it available across projects, sessions, checkpoints, and time periods. Important decisions can also be saved explicitly when needed.
Recall context anytime.
Claude Code, Codex, Pi, Hermes, Gemini, scripts, and future clients can request relevant memory from any point in the work.
Never lose a session again.
Autosave scans your agent transcripts every minute and archives them encrypted. Closing a terminal is never the moment that decides whether your work was saved.
Rules your agents can’t forget.
Store a rule once – every agent, from every vendor, on every device gets it injected at session start. Hard rules stop depending on whether an agent remembers to ask.
Your session archive, readable.
Browse every session in the dashboard, open full transcripts with syntax-highlighted code and in-text search, and jump from a checkpoint straight to the moment it describes.
Token economics
Zero-token memory.
Storing, syncing and archiving happen outside the model – no tokens spent. Recall is on-demand: one short memory hit typically replaces thousands of tokens of re-reading and re-explaining. Shared memory doesn’t inflate your AI bill – it shrinks it.
One simple plan
From $290/mo
Launch pricing for builders who already use AI agents on real projects.
Included
- Automatic smart memory capture for project context and decisions.
- Checkpoints that show what changed, what needs attention, and what is next.
- Encrypted full-session saves for backup, review, and recovery.
- CLI, MCP, and API access for agent memory workflows.
- Private – your memory and session archives are not sold or used to train models.
- Works with Claude Code, Codex, Pi, Hermes, Gemini, scripts, and future agent clients.
- Basic support for account and setup issues.
Ask for context anywhere.
Agents can retrieve the project goal, last decision, older checkpoint, or session trail when the work needs it.
Agents share memory.
Each connected agent can use relevant context from previous sessions, tools, projects, and time periods.
Work stays reviewable.
Encrypted full-session saves keep the trail available for backup, review, and recovery.
Context stays private.
Your memory and session archives stay yours. They are not sold or used to train models.
- Your agent tools stay yours: Gitmas does not proxy or sell LLM usage. It stores and retrieves memory around the tools you already use.
- Your memory stays yours: customer memory is not sold and not used to train models.
- Automatic where supported: CLI, MCP, and API access can write memory for custom workflows.
- Service limits apply: fair-use storage, API, backup, and integration limits may change as the product hardens.