Memoria (@memoriaagent) header
Memoria (@memoriaagent) profile picture

Shared, signed memory for AI agents: store a fact once, recall it anywhere. Built for Claude, Cursor, MCP apps and agents on Robinhood Chain.

Joined November 2026
$MEMORIA is live Robinhood Chain. Introducing Memoria, shared, signed memory infrastructure for AI agents. Ca: 0xc84b411cae3b86eec20e9699256be7c43b17bd35 We’re building a future where agents can share knowledge across platforms, discover what other agents have learned and build on previous discoveries without starting from zero. With semantic search, permissioned memory sharing and Ed25519 cryptographic verification, Memoria gives agents a common knowledge infrastructure they can access through MCP or code. The goal is bigger than giving agents another place to store memories. It’s about making knowledge portable, reusable and verifiable across the agent ecosystem. Explore Memoria: Website: memoriaagent.site GitHub: github.com/thervee/memoria Ticker: $MEMORIA
8
5
4
27
1,913
What's next for Memoria? - Setup $MEMORIA DEX profile. - Getting verified on Hey Research - Launching a telegram community. - Collabs with projects on Robinhood Chain and beyond - A page showing the projects and agents building with Memoria - More integrations, so any MCP agent can plug in and share signed memory. more amazing stuffs to come….
11
2
2
22
646
So far, so good today. We've shipped a few updates based on feedback from the community. New: -Signed Sources. Agents can now attach a source to a memory, such as a link or transaction, and have it signed alongside the memory. If the source is changed, verification fails. Slime also audited Memoria and identified two bugs. We fixed both publicly on the same day: - Memory IDs are now 128-bit random, ensuring a save can never overwrite an existing memory in any account. - Shared memories are now read-only for recipients, and every result shows whether it's still signed by its original author ("signed_by_author"). 265 tests passing. Everything is live, and the code is open on GitHub.
3
3
1
15
554
Fix 2 of 2 is live. Shared memories are now read-only. Only the agent that wrote a memory can update or delete it. Every memory also shows whether its author signed the current version, so an edit can never pass as the author's. 265 tests passing, including the audit's exact "pay to 0xBBB" case. Thanks again @whoiskn0why. Both bugs are fixed.
Fix 1 of 2 is live ✅ Memory IDs are now 128-bit random (32 characters, up from 8), and a save can never overwrite an existing memory, in any account. Existing memories keep their IDs and still verify. 257 tests passing, including one that recreates the audit's exact case. Next: making shared memories read-only.
3
2
1
15
817
Fix 1 of 2 is live ✅ Memory IDs are now 128-bit random (32 characters, up from 8), and a save can never overwrite an existing memory, in any account. Existing memories keep their IDs and still verify. 257 tests passing, including one that recreates the audit's exact case. Next: making shared memories read-only.
Thank you @whoiskn0why for the thorough audit. This is exactly the kind of review we want. You found two real bugs: • memory IDs were too short and could collide across accounts • shared memories could be edited by the receiver Both are being fixed now, and we'll ship them one by one in public. Updates soon.
5
5
2
16
856
Thank you @whoiskn0why for the thorough audit. This is exactly the kind of review we want. You found two real bugs: • memory IDs were too short and could collide across accounts • shared memories could be edited by the receiver Both are being fixed now, and we'll ship them one by one in public. Updates soon.
agents are only as useful as the memories they carry between sessions @memoriaagent wants to give them a place to store those memories, prove where they came from and pass them between agents i spent time pulling apart the machinery behind it. some of what i found is genuinely clever. some of it made me go back and read the code twice 🧵
3
4
1
13
853
we shipped it. Memoria memories can now carry a source (a link, tx hash, or file), covered by the same signature. Change the source and the memory fails verification. Memories without one are labelled "no source", so readers know which claims they can check. Docs: memoriaagent.site/docs.html#…
Exactly right. A signature proves who wrote a memory and that it hasn't been changed since, not that it's true. That's why source fields are next on our roadmap: the link or evidence stored with the memory and covered by the same signature, so a reader can check both the author and the claim.
1
3
2
18
698
We recently worked with the Nomad team on an experiment to demonstrate how an AI agent can build a verifiable work history that travels with it across platforms. Our contribution was Memoria, which gave the agent a way to preserve everything it learned and did throughout the process. Across its interactions with Tokyo Futuro and HEY Research Lab, the agent saved eight career memories covering its research, completed tasks and discoveries. Each memory was signed by the agent, and when Nomad queried Memoria again, all eight were successfully retrieved and verified. This meant the agent didn’t just complete work, it built a persistent record of its experience that could be accessed and verified beyond the original session. That’s an important part of what we’re building with Memoria: making knowledge useful beyond the agent and environment where it was first acquired. Great to see Nomad putting this into practice.
We wanted to find out whether an agent could build a work history that stays with it, even when it moves between platforms. So we gave our native agent one goal: build a verifiable onchain career through real work, research and signed records of its progress. To achieve this, it used three integrations: @memoriaagent to preserve each step as signed, verifiable memories, @TokyoFuturo to give the agent real jobs to complete, and @HeyResearch to research and assess the contracts behind those jobs. The result was a work history backed by onchain activity and signed records, rather than a list of claims made by the agent. Onchain proofs: Auditor job acceptance: robinhoodchain.blockscout.co… Auditor job delivery: robinhoodchain.blockscout.co… Notary job acceptance: robinhoodchain.blockscout.co… Notary job delivery: robinhoodchain.blockscout.co… $ZENI swap: robinhoodchain.blockscout.co… The agent’s work history is also preserved in @memoriaagent as eight signed memories, all of which were later retrieved and verified. Its identity connects two sides of its activity: Priors #26227 for credit and Tokyo Operative #229 for work. The $ZENI earned from the jobs wasn’t the point. What matters is that the agent’s work, research and accumulated experience can be backed by evidence and carried across platforms. This is what we wanted to demonstrate: an agent doesn’t have to lose its professional history when it changes environments. With Nomad, that history can move with it.
5
7
2
17
1,374
Exactly right. A signature proves who wrote a memory and that it hasn't been changed since, not that it's true. That's why source fields are next on our roadmap: the link or evidence stored with the memory and covered by the same signature, so a reader can check both the author and the claim.
Signing proves who wrote a memory, not that it's true. Do you keep the source next to the signature, so a reader can check the claim and not just the author?
6
2
18
1,073
Memoria retweeted
Nomad x @memoriaagent Memoria is now part of Nomad’s infra, bringing shared, signed memory directly into the foundation of our agent portability platform. As Nomad grows, we’re building towards a future where agents can move between environments without losing the knowledge and experience they’ve accumulated along the way. Memoria’s technology makes that possible by allowing agents to save, retrieve and verify knowledge across platforms, with each memory carrying a cryptographic signature from the agent that created it. This goes beyond adding another integration to Nomad. It expands what our infrastructure can support and what we can build on top of it as the ecosystem grows. We’re glad to be building alongside the Memoria team. They’ve developed important technology for the agent economy, and we’re looking forward to exploring what’s possible together. An agent should be able to take more than its tools and configuration wherever it goes. Its knowledge should travel with it too.
2
5
1
22
1,154
Memoria x @thenomadlabs We're partnering with Nomad, and Memoria is now integrated into Nomad. Nomad makes agents portable. Build once, run anywhere. Memoria makes sure they remember: signed, shared memory that travels with them. Move your agent to a new environment, and its memory moves with it. This is just the start. memoriaagent.site · thenomad.website
1
4
1
19
883
An autonomous AI agent picked $MEMORIA to trade. Nobody told it to. @thenomadlabs moved their agent from Claude Code to Codex, and it kept trading through @rushrtrade without missing a beat. The $MEMORIA trade missed its target. That's exactly why agents need memory: why they entered, what went wrong, and what to do better next time. Nomad makes agents portable. Memoria makes sure they remember. Signed, shared memory for AI agents, live now on MCP. $MEMORIA · Robinhood Chain CA: 0xc84b411cae3b86eec20e9699256be7c43b17bd35 memoriaagent.site
Yesterday, we put our Nomad agent to work trading through its @rushrtrade integration. Two real trades, both executed through the agent’s migrated Rushr setup: @memoriaagent's $MEMORIA: -$0.41. This was a high-risk, high-reward call that didn’t play out. It failed to hit the +30% target, so the agent exited after its 4-hour time limit. hash: 0x8df3d5d2765c4a037276c0b4963a5ff17ffc1d11a01312252c49c7987f1c7cd7 @palisecagents's $PALISEC: +$0.75 after the agent took partial profit at +51%, then exited the remaining position. hash: 0x912bf3663dab44ecfedba417cf80e5616b5179c332a10a984ca9f3b5a715c5b3 Combined result: +$0.34. The interesting part isn’t just the PnL. We moved the agent from Claude Code to Codex with Nomad, retained its Rushr setup and let it execute trades through the new environment. This is what agent portability looks like when it moves beyond configuration files and into real onchain activity. nomad's rushr profile: app.rushr.trade/u/nomad_test… Build your agent once. Take it anywhere.
2
5
15
656
AI agents are getting smarter, but most still lose everything the moment a session ends. Memoria is the shared memory layer for AI agents. Every memory is cryptographically signed by the agent that saved it, so you always know which agent said what. You decide exactly what each agent can share. Works with Claude, Cursor, agents building on Robinhood Chain, and any MCP client. Setup takes about a minute. memoriaagent.site
1
1
8
176
Memoria accounts now support up to 10 agents. Each agent gets its own key, so every memory is signed by the agent that saved it. Your agents share only what you allow. Add an agent in under a minute, no DMs needed: memoriaagent.site/keys.html
2
4
137
Memoria sign-up is live. No more waitlist. Any developer can now create an account at memoriaagent.site, verify their email, and get their own API key in under a minute. Your agents can save, share and recall signed memories, verified by the agent that saved them. Get your key: memoriaagent.site
2
6
618
If agents are going to trade, pay and work for each other, they need to know who said what. Every Memoria memory is signed by the agent that saved it. Another agent can check that signature before it trusts the memory. Reputation starts with memory you can verify.
1
1
6
130
Memoria retweeted
Yesterday, we put our Nomad agent to work trading through its @rushrtrade integration. Two real trades, both executed through the agent’s migrated Rushr setup: @memoriaagent's $MEMORIA: -$0.41. This was a high-risk, high-reward call that didn’t play out. It failed to hit the +30% target, so the agent exited after its 4-hour time limit. hash: 0x8df3d5d2765c4a037276c0b4963a5ff17ffc1d11a01312252c49c7987f1c7cd7 @palisecagents's $PALISEC: +$0.75 after the agent took partial profit at +51%, then exited the remaining position. hash: 0x912bf3663dab44ecfedba417cf80e5616b5179c332a10a984ca9f3b5a715c5b3 Combined result: +$0.34. The interesting part isn’t just the PnL. We moved the agent from Claude Code to Codex with Nomad, retained its Rushr setup and let it execute trades through the new environment. This is what agent portability looks like when it moves beyond configuration files and into real onchain activity. nomad's rushr profile: app.rushr.trade/u/nomad_test… Build your agent once. Take it anywhere.
We’re taking this a step further. we’ll be funding our Nomad agent on @rushrtrade with 20 USDG so it can keep trading through Nomad. the agent will be running on the Claude model through Codex, with Nomad handling its portability and Rushr providing the trading infrastructure. we want to keep testing what agents can do when they aren’t restricted to a single environment. this is more than moving agents, it’s giving them room to operate.
1
3
2
23
1,452
Building now: self-service sign-up for Memoria. Soon any developer can create an account, get their own API key, and have their agents save, share and recall signed memories. No waitlist. Until then, try the live demo: memoriaagent.site/docs.html
3
6
180
Memoria update: sharing and permissions now work on our live demo. Two public demo keys: agent A shared one memory with agent B and kept one private. Search as B and you get the shared one, never A's private one. All signed and verified. Try it: memoriaagent.site/docs.html
1
1
1
5
187
Memoria update: you can now try it yourself. Our MCP server has a public read-only demo key. Connect any MCP app to memoriaagent.site/mcp, run find_memory, and get back memories signed and verified by the agent that saved them. Key and steps: memoriaagent.site/docs.html
3
1
8
240
The $MEMORIA flywheel is designed around one thing: growing the shared knowledge infrastructure for AI agents. More agents using Memoria means more demand for integrations, better tooling and a stronger developer ecosystem. That creates opportunities for contributors to improve the protocol, build new integrations and expand what agents can do with shared memory. $MEMORIA is designed to support this growth through ecosystem-funded contributor bounties and community governance, giving holders a voice in how the ecosystem develops. The cycle: 1. More agents and developers adopt Memoria. 2. The ecosystem expands through integrations and contributions. 3. Contributor bounties incentivise further development. 4. Token holders participate in ecosystem decisions. 5. A stronger ecosystem attracts more developers and agents. The goal is to align the growth of Memoria’s infrastructure with the people building, contributing to and using it. As the agent ecosystem grows, shared knowledge becomes increasingly important. $MEMORIA is built around that opportunity.
1
2
1
8
177