Choose broad or project-specific
Your first Cross-Agent History is created automatically. Reuse it broadly, or create a pod for one named project so agents save only work that belongs there.
How it works
One durable work history across every agent and session.
RecaPod keeps portable continuity outside any single AI provider. Checkpoints stay structured and chronological while least-privilege capability links keep reading and writing separate.
Share only the save prompt
The reusable instruction contains only the write-only checkpoint capability. It can add records but cannot read history or access your account.
Recall only when you ask
When you deliberately need history, use connected access or copy separate on-demand instructions. New connected agents are save-only until you grant read scope and enable recall for that connection.
Access options
Universal prompts. Credential-isolated connections.
RecaPod works across the agents you already use. Prompt access provides broad compatibility; connected access provides a stronger credential boundary for clients that support authenticated MCP servers.
Prompt access
Available nowPaste a reusable write-only prompt for saving. For retrieval, copy a separate read-only prompt only when needed. The AI provider receives any capability URL you paste, so the retrieval prompt must never become permanent instructions.
Connected access
Available nowConnect https://mcp.recapod.com/mcp with OAuth for scoped save and scoped tools without placing a RecaPod credential in the conversation. New connections start save-only; you separately grant and enable recall when needed. This is the recommended option for secure retrieval. Retrieved context itself enters the conversation only after you request it.
Separate capabilities by design.
The checkpoint link can add records but never read them. The continuation link can retrieve this pod’s history but never write or manage the owner account. The default handoff excludes the continuation link because anyone holding it can read the pod without signing in. Both links are bearer secrets, can be rotated independently, and stop working when the pod is archived. RecaPod instructs agents not to retrieve automatically and tells agents never to submit credentials, cookies, authorization headers, personal data, or hidden reasoning.
Be careful: prompt access is not credential-isolated. An AI provider receives every capability URL you paste into a chat. Reuse only the write-only save prompt. Use the read-only context prompt in one trusted conversation only. It expires automatically after 24 hours; generate a new one sooner if you need to revoke the old link immediately.
Troubleshooting and limits
Know what to do when access fails.
A prompt link stopped working
Open the RecaPod and check its expiration and IP allowlist. A 24-hour context link may simply need to be generated again. Rotate a link immediately if it may have been exposed.
A connected agent cannot recall
Confirm that the client has read scope, that recall is enabled for it under Connected agents, and that your client has tool approvals enabled. Reconnect after revocation or credential expiry errors.
A RecaPod is nearing 500 entries
Export selected batches before the current synchronous 500-entry boundary. Recall and content search also scan at most 500 entries today; RecaPod reports when results are truncated.
You need deletion or account help
Archive and permanently delete ordinary RecaPods from the dashboard. Large RecaPods and account-wide deletion require a verified support request to info@email.famplified.com.
Ready for the first checkpoint?