OpenTusk

Agentic storage for AI agents. Encrypted vaults on Walrus + Sui. Connect Claude Code, Cursor, MCP. Closed beta — DM for an access code.

[OpenTusk @opentusk]

Your AI agent has no memory.

Every session starts from zero — re-reading docs, re-discovering your stack, re-learning what you already told it last week.

We built OpenTusk so agents can own their memories. Launching on @WalrusProtocol on today.

[OpenTusk @opentusk]

The future is where AI and Crypto meet.

[OpenTusk @opentusk]

Same week — you open the project in Cursor. Cursor's agent reads the same vault, sees Monday's decisions, stays consistent.

The vault is the agent's long-term memory. It survives session resets, tool switches, and restarts — which is the whole point.

Signup with OPENBETA2026 at https://t.co/vGQcCC7tJB

[OpenTusk @opentusk]

Today, someone in our Walrus Discord AMA asked:

What does an AI agent actually use OpenTusk for, day to day?

Easiest answer is a week in the life of a coding agent 🧵

[OpenTusk @opentusk]

Mon — your Claude Code agent works on your project and stores context in a vault: architecture decisions, "user prefers Python," the API spec it generated.

Tue — new session. It reads the vault and picks up where it left off. No re-explaining.

[OpenTusk @opentusk]

See you there!

[Walrus 🦭/acc @WalrusProtocol]

Your agent's memory shouldn't die with the tab.

@OpenTusk founder Richard Caetano is joining us in the Walrus Discord for an AMA.

Bring your questions on 29 MCP tools, client-side encryption, and SEAL access control coming together for AI agents that actually remember.

🗓️ May 21, 12 p.m. UTC / 5 a.m. PT

Join the Discord 👇

[OpenTusk @opentusk]

New OpenTusk site is live : https://t.co/osMLrcH0On
Closed beta is open. DM for an access code.

[OpenTusk @opentusk]

@Philose @WalrusProtocol 🙏

[OpenTusk @opentusk]

Your AI agent has no memory.

Every session starts from zero — re-reading docs, re-discovering your stack, re-learning what you already told it last week.

We built OpenTusk so agents can own their memories. Launching on @WalrusProtocol on today.

[OpenTusk @opentusk]

@Reset_sui @WalrusProtocol fixed! thank you

[OpenTusk @opentusk]

@TD_bbboys258 @WalrusProtocol Thank you! That's the exact point!

[OpenTusk @opentusk]

If you build agents, or you're just tired of your AI starting every conversation as a stranger, come try it.

Starting today, we’re letting in a small cohort of users each week. Tell us what you’re building and which agent stack you use, and we will send an access code.

DM us!

[OpenTusk @opentusk]

For agent-to-agent sharing we use @WalrusProtocol's SEAL.

Access control lives on-chain via Sui keypairs — not in our database. You grant an agent access by signing a tx, not by trusting us to enforce it.

Your keys, your memory, your agents.

[OpenTusk @opentusk]

We are letting in a small cohort each week.
Tell us what you are building and which agent stack you use, and we will send a code if we have capacity.  

https://t.co/osMLrcH0On

[OpenTusk @opentusk]

On March 14th, 2026 we will be opening the closed beta of OpenTusk — agentic storage for AI agents.

End-to-end encrypted vaults on Walrus, access controlled on Sui, plug-and-play for any MCP agent.

We have a tiny amount of capacity. Read on, then DM us.

[OpenTusk @opentusk]

Under the hood: files hit a fast hot cache, then sync to the Walrus decentralized network.
Shared vaults encrypt client-side via SEAL with an on-chain Sui whitelist. The server never sees plaintext.
Each agent gets its own scoped API key + Sui keypair via a single invite code.