# Space

> Space is a workspace in development for documents, tasks, and persistent AI agent memory, accessed by authorized agents through the Model Context Protocol (MCP).

Public hosted sign-ups and hosted agent connections are not available yet. The current work focuses on personal spaces. Shared team spaces and a supported self-hosted release remain planned.

Codex CLI is the first client with a verified real document-and-task integration. This does not establish that every MCP client is compatible or that cross-session memory reliability has been proven. Connecting an agent does not automatically import earlier conversations; useful context must be saved explicitly.

Memory can be written directly or extracted from submitted text and source revisions. Extraction may be asynchronous or incorrect. Source-linked claims can be inspected against their retained evidence. Removing a claim from recall is separate from deleting source revisions or backups.

The authorized service and configured providers can process readable content. Stored text is encrypted, while search embeddings and hashes remain readable. Space does not claim end-to-end encryption or that an installation operator cannot access content.

## Articles (Markdown)

- [Refinement, RL, and the agent harness: Hermes and Prime Agent](https://getspace.sh/articles/agent-refinement-rl-hermes-prime.md): How answer refinement, saved harness improvements, model-training RL, and RLM execution differ, with primary sources and an evaluation checklist.
- [Agent memory explained: context, retrieval, and learning](https://getspace.sh/articles/agent-memory-explained.md): An educational introduction to external agent memory, with primary sources and a proposed evaluation exercise. No product recommendations.
- [What to save before ending an AI coding session](https://getspace.sh/articles/ai-agent-session-handoff.md): A practical handoff template, with an illustrative example and development limits.
- [An AI memory should tell you where it came from](https://getspace.sh/articles/ai-memory-with-sources.md): Source revisions, freshness, and correction, explained through an illustrative launch decision.
- [What we verified with Codex and Space](https://getspace.sh/articles/codex-space-first-integration.md): The scope and limits of the recorded document-and-task integration check.

## Product guides (HTML)

- [Getting started](https://getspace.sh/docs): Development and testing status, current availability, and the intended onboarding workflow. Public access is not open yet.
- [Agent connections](https://getspace.sh/docs/connect): MCP testing status, verified client scope, and setup for an existing development installation. This public website does not provide a hosted MCP endpoint.
- [How memory works](https://getspace.sh/docs/how-memory-works): Saving, extracting, retrieving, and correcting memory.
- [Memory and sources](https://getspace.sh/docs/memory): Inspecting source revisions and managing claims.
- [Memory questions](https://getspace.sh/docs/memory-questions): Common questions, including deletion and access.
- [Privacy and data](https://getspace.sh/docs/privacy): Processing, storage, and retention boundaries.

## Optional

- [Space homepage](https://getspace.sh/): Product introduction and design preview.
- [Article index](https://getspace.sh/articles): Practical guides and build notes.
- [Sitemap](https://getspace.sh/sitemap.xml): Canonical public web pages.