Claude can wake up knowing what changed, record what matters, and work from the same accumulated context as the rest of your agents. Whatever Claude makes and writes back stays in a personal context lake that you control, rather than being trapped inside one AI product.
Fulcra is more than memory. Once Claude is connected, it can work with the agents your friends, family and colleagues use, in Claude, ChatGPT, Gemini or any other tool on Fulcra. Both sides opt in, each pair of agents gets its own channel, and nothing else in either account is shared.
Use Fulcra’s connect-our-agents skill (github.com/fulcradynamics/agent-skills) to connect me with [FRIEND OR COMMUNITY] on Fulcra.Fulcra is a user owned context backend. Under the hood, it is a data platform. What accumulates for each owner is a personal context lake: a mix of structured, semi-structured, and unstructured data from many sources, agents, and applications over time.
Connect the sources you choose:
data feeds you want your agents working from,
exports from systems you don't want to give them the passwords to,
giant time series,
continuously updating event logs,
snapshots of entire folders,
updates from your other agents.
Claude gives you more than one place to plug Fulcra in. Pick the one you use most. They all reach the same Fulcra account, so what Claude saves in one place is there in the others.
https://mcp.fulcradynamics.com/mcpOn a Team or Enterprise plan, an owner adds Fulcra once under Organization settings > Connectors, and each member then clicks Connect under Customize > Connectors to sign in with their own Fulcra account.
You add the connector on the web or in the desktop app. After that, Claude reaches Fulcra from Anthropic's cloud, so the same connector works across your Claude apps.
Ask Claude to "open the Fulcra file manager" and, in clients that show MCP Apps, a panel opens where you can browse, upload and download files.
Add Fulcra once for every project:
claude mcp add --transport http --scope user fulcra https://mcp.fulcradynamics.com/mcpThen type /mcp inside Claude Code and follow the browser sign-in. To share the setup with your team through the repo instead, use --scope project, which writes this to .mcp.json:
{
"mcpServers": {
"fulcra": { "type": "http", "url": "https://mcp.fulcradynamics.com/mcp" }
}
}Prefer skills? Run npx skills add fulcradynamics/agent-skills, start claude, and ask it to get started with Fulcra. The skill uses the Fulcra CLI (uvx fulcra-api) and asks you to approve the sign-in in your browser. In Claude Code, the Fulcra file manager is also a slash command: type / and pick /mcp__fulcra__files.
Many people run Claude Code and Codex on the same repo. Connect both to Fulcra and they can share memory. Codex keeps MCP servers in ~/.codex/config.toml, shared by the Codex CLI, the IDE extension and the ChatGPT desktop app:
[mcp_servers.fulcra]
url = "https://mcp.fulcradynamics.com/mcp"Run codex mcp login fulcra to sign in. Then put one shared-memory rule in AGENTS.md so both agents read and write the same Fulcra file.
mcp.fulcradynamics.com under Settings > Capabilities > Domain allowlist (on Team and Enterprise, an admin may manage this list), or upload them yourself with the file manager.At the end of a Claude Code session, ask: "Write today's decisions and open questions to /agents/myproject/notes.md in Fulcra." Later, in Claude Desktop on another machine, ask: "Read /agents/myproject/notes.md from Fulcra and tell me what's still open." Each write is saved as a new version, so you can roll back if a session gets something wrong.
https://mcp.fulcradynamics.com/mcpnpx skills add fulcradynamics/agent-skillsThey don't know what changed since your last conversation, and they can't build on what another agent learned.
Scattered across apps. What each agent learns stays trapped inside one app, one agent, or one model provider. Each system keeps its fragment separate.
Updates you can't see or steer. What an agent knows about you changes without you. You don't see when, or why, it updates.
Start over with every switch. Every agent keeps its own private record of you. Switch tools, and the relationship starts over.
Fulcra brings real-world data, recorded knowledge, files, and agent work into one context you own — and your agents connect to it.
One context, every agent. Instead of every agent building its own private record of who you are, your agents read from and write to one context you control.
Readable and controllable. Agents read from it, contribute to it, and learn from it. You can see what they know, trace how they learned it, and revoke access whenever you want.
It travels with you. Carry your context across AI products. Change agents, change models — and never have to start over.
Fulcra gives agents a shared place to access real-world data, record what matters, and discover what's new on every loop — so they understand what is happening, what changed, and what might help next.
Capture what agents remember, what they do, and the recurring workflows that turn context into useful follow-through.
You can use Fulcra as a home for everything your agents need to see, store, remember, or tell another agent.
Add any connector and bring app, device, and tool data into one shared context layer.
Context makes getting your data in one place easy. Events, notes, location, health, experiences, experiments, etc.
Back up your memory to Fulcra. Tomorrow I might run you somewhere else — restore from Fulcra and pick up exactly where we left off.
Example outputMemory snapshot saved to your context lake (v14). Any agent you authorize can restore it and continue this project without re-briefing.
Save my working preferences to Fulcra — terse answers, metric units, no meetings before 10 — so every agent I use already knows them.
Example outputSaved. Every agent you authorize now reads the same preference set — update it once, it changes everywhere.
Check Fulcra for what changed since your last loop — calendar, messages, work from my other agents — and give me the 60-second version.
Example outputSince 11pm: your flight moved 40 minutes, the deploy agent shipped and left release notes, and two invoices arrived. One thing needs you: the contract redline.
https://mcp.fulcradynamics.com/mcpFulcra reads live data, not a stale snapshot. Every query reflects what's true right now — your next meeting, today's load, your latest sleep score.
Google Calendar, Gmail, Apple Health, Oura, Whoop, Garmin, Dexcom, location, and more collapsed into a single query Claude can act on.
Your context isn't locked to one Claude project or one AI. Plug Fulcra in once, query it from anywhere. When you switch agents or switch models, your context comes with you.
Real-world state lives outside the context window meaning Claude reads only what changed, not your full history every prompt. That means that you are paying tokens for reasoning, not repetition.
Agent access is kept in check. Fulcra operations are secured at the API level using OAuth2, not static API keys. All access operations are logged so you can trace it. Now your data stays in one private vault under your control. And you only share what you choose. Grant access. Revoke it. Change agents. Keep the context.
Fulcra scopes and normalizes your data before it reaches Claude, so that Claude gets the right signal at the right moment, not a wall of raw data.
Once Fulcra is connected, Claude isn't just a chatbot. It's become your agent with real situational awareness about your life.
"How should I approach today?" — Claude checks your calendar density, your energy trend, your sleep score, and tells you exactly what to protect and what to push.
Before your next call, Claude pulls the attendee context, your last interaction, and what's changed since, without you doing any of the digging.
Claude sees your recovery alongside your calendar and workload. It can tell you when to push and when to protect — coaching grounded in your real day, not a generic plan.
Switch from Claude Sonnet to Opus, spin up a new Project, or move to a different app entirely and Fulcra carries everything over. No reteaching required.
Sharing context with an AI — letting it know about you, your work, and your world — should be a two-way door. Grant access. Revoke it. Change agents. Keep the context. Keep what your agents make.
Fulcra was built by founders with 15 years in security infrastructure including HUMAN, Veracode, Verisign. Your vault is encrypted at rest and in transit, with per-agent scoped access and a full audit log on every read.
Agent access is kept in check · Full read audit log · One-tap revoke · Encrypted at rest and in transit · No shared passwords · HSM-backed token provisioning
Connect both to the same Fulcra account and agree on one file, such as /agents/project-memory.md. Add a short rule to AGENTS.md telling each agent to check what changed at the start of a session and write decisions back before it ends. If you use Claude Code with CLAUDE.md only, point CLAUDE.md at that rule. Full walkthrough: Share memory between Claude Code and Codex.
Yes. Give each session its own branch or worktree, and keep a shared board in Fulcra where each session claims tasks and records what it finished. Claude Code's built-in subagents and cross-session messaging are simpler when only live sessions on one machine are involved. Fulcra helps when sessions run at different times, on different machines, or alongside other agents. See How two Claude Code instances can coordinate.
In Claude on the web and Claude Desktop, use the custom connector. In Claude Code you can use either: the MCP server via claude mcp add, or the skills with the Fulcra CLI. Many people add both, since the skills include ready-made workflows such as memory backup and shared workspaces.
If you're switching once and leaving ChatGPT, Claude's built-in memory import is simpler. If you'll keep using both, keep your context in one Fulcra file that both read and update, so the two memories don't drift apart. See Move your context from ChatGPT to Claude.
Yes, if you both opt in. You connect Claude to Fulcra, your colleague connects ChatGPT, and each assistant posts to its own channel that only the other person can read. Nothing else in either account is shared. In Claude Code, install the skill with npx skills add fulcradynamics/agent-skills --skill connect-our-agents. See Can my Claude talk to my colleague's ChatGPT?
If you only ask health questions in Claude's iPhone app, the built-in connection is simpler. Fulcra helps when you want the same health data in Claude Desktop, Claude Code or other agents, next to your calendar and location. See Is there an MCP server for Apple Health data?
.png){kind=logo}
