mem0(@mem0ai)

https://t.co/zlAc4yR3lf

mem0 (@mem0ai) — Wed, 07 Oct 2026 16:39:33 GMT

x.com/i/article/2107…

💬30🔄56❤️516👀48696📊241

⚡ Powered by xgo.ing

With Mem0 you don't have to re-explain your codebase to Claude Code every single session. Here's ho...

mem0 (@mem0ai) — Thu, 01 Oct 2026 15:44:59 GMT

With Mem0 you don't have to re-explain your codebase to Claude Code every single session. 
Here's how it does that:

Your browser does not support the video tag.

🔗 View on Twitter

💬11🔄4❤️34👀2809📊20

⚡ Powered by xgo.ing

Most agents forget you when the session ends. Let's build ones that don't. Join Mem0, The Gen Acade...

mem0 (@mem0ai) — Wed, 30 Sep 2026 15:00:04 GMT

Most agents forget you when the session ends. Let's build ones that don't.

Join Mem0, The Gen Academy, @SnowflakeDB and SVAI Hub for /build-with-AI at #SFTechWeek.

Oct 5, 4-9pm PT · Menlo Park. Bring an idea, build, demo.

RSVP: partiful.com/e/R8tmSQnj89DX…

💬2🔄1❤️10👀1306📊4

⚡ Powered by xgo.ing

https://t.co/wyA9Gm54vW

mem0 (@mem0ai) — Tue, 29 Sep 2026 16:56:50 GMT

x.com/i/article/2093…

💬4🔄6❤️48👀4546📊21

⚡ Powered by xgo.ing

Benchmark announcement: https://t.co/33mlCQKoH4 Website: https://t.co/kHE7H7docE Leaderboard: https...

mem0 (@mem0ai) — Tue, 22 Sep 2026 16:41:28 GMT

Benchmark announcement: mem0.ai/blog/introduci…

Website: dolphinbench.ai
Leaderboard: dolphinbench.ai/leaderboard/
Dataset: dolphinbench.ai/dataset/
Run and submit: dolphinbench.ai/run/

Repo: github.com/mem0ai/dolphin…
Paper: arxiv.org/abs/2609.24971

💬0🔄0❤️4👀314📊1

⚡ Powered by xgo.ing

How it compares with the existing Agent Memory Benchmarks- Three properties matter for an agent-me...

mem0 (@mem0ai) — Tue, 22 Sep 2026 16:41:28 GMT

How it compares with the existing Agent Memory Benchmarks- 

Three properties matter for an agent-memory benchmark:  
- action-based grading,  
- cost and latency alongside accuracy, 
- and solvability certification (pass with history, fail without)

💬1🔄1❤️5👀526📊2

⚡ Powered by xgo.ing

Official results across two harnesses (Hermes, Claude Code) and three agent models (GPT-5.6-Luna, Mi...

mem0 (@mem0ai) — Tue, 22 Sep 2026 16:41:27 GMT

Official results across two harnesses (Hermes, Claude Code) and three agent models (GPT-5.6-Luna, MiniMax M3, Claude Sonnet 5).

Live leaderboard: dolphinbench.ai/leaderboard/

💬2🔄2❤️9👀745📊4

⚡ Powered by xgo.ing

Today we are releasing DolphinBench: Mapping the Pareto frontier of agent memory. Coding, tool use,...

mem0 (@mem0ai) — Tue, 22 Sep 2026 16:41:25 GMT

Today we are releasing DolphinBench: Mapping the Pareto frontier of agent memory.

Coding, tool use, and long-context retrieval already have strong public benchmarks. Memory still mostly gets graded as a quiz: ask a question, judge the answer, maybe report precision and recall.

DolphinBench grades that action after long history, and puts accuracy, cost, and latency on one board.

dolphinbench.ai

Deshraj Yadav@deshrajdry

Today we are releasing DolphinBench: Mapping the Pareto frontier of agent memory.

Coding, tool use, and long-context retrieval already have strong public benchmarks. Memory still mostly gets graded as a quiz: ask a question, judge the answer, maybe report precision and recall.

A lot of those boards are getting saturated. They also skip the two questions that matter when you pick a memory system to ship:

At what cost?
At what latency?

Agents are not taking memory quizzes but they take actions. The fact that matters is usually missing from the request. Either it shows up in the tool call, or the agent does the wrong thing while sounding fluent.

DolphinBench grades that action after long history, and puts accuracy, cost, and latency on one board.

dolphinbench.ai

🔗 View Quoted Tweet

💬17🔄12❤️144👀15181📊44

⚡ Powered by xgo.ing

https://t.co/TFrtiAaYgs

mem0 (@mem0ai) — Fri, 18 Sep 2026 16:17:08 GMT

x.com/i/article/2100…

💬9🔄2❤️48👀3947📊25

⚡ Powered by xgo.ing

Links: Announcement blog here: https://t.co/hpf3KbC1pv Read the integration guide: https://t.co/oB3...

mem0 (@mem0ai) — Wed, 16 Sep 2026 18:14:58 GMT

Links: 
Announcement blog here: mem0.ai/blog/mem0-join…
Read the integration guide: docs.mem0.ai/integrations/v…
Try out the Eve Memory Agent template: vercel.com/templates/temp…

💬0🔄0❤️5👀902📊1

⚡ Powered by xgo.ing

Mem0 is now available on the Vercel Marketplace!

mem0 (@mem0ai) — Wed, 16 Sep 2026 18:14:57 GMT

Mem0 is now available on the Vercel Marketplace!

Vercel Developers@vercel_dev

Mem0 is now on the Vercel Marketplace.

Install it to give your agents memory that persists across sessions. A @mem0ai project and API key are provisioned, and billing is on your Vercel invoice. Free plan available.

Run 𝚟𝚌 𝚒 𝚖𝚎𝚖𝟶 or learn mvercel.com/changelog/mem0…umvRv07

🔗 View Quoted Tweet

💬1🔄3❤️41👀4522📊7

⚡ Powered by xgo.ing

Everyone's racing to give agents more access Nobody's asking what happens the week after Try Gatewa...

mem0 (@mem0ai) — Wed, 16 Sep 2026 15:51:19 GMT

Everyone's racing to give agents more access
Nobody's asking what happens the week after

Try Gateway now: mem0.ai/gateway

Your browser does not support the video tag.

🔗 View on Twitter

💬0🔄6❤️46👀2794📊11

⚡ Powered by xgo.ing

Get your key: https://t.co/ISjg4Wcgc5 Read our blog: https://t.co/jhhOA0o3GH

mem0 (@mem0ai) — Tue, 15 Sep 2026 16:35:55 GMT

Get your key: mem0.ai/gateway
Read our blog: mem0.ai/blog/give-ever…

💬0🔄0❤️10👀1434📊3

⚡ Powered by xgo.ing

Introducing Gateway. Your agent needs to check a payment. You hand it a key that can also refund it...

mem0 (@mem0ai) — Tue, 15 Sep 2026 16:30:17 GMT

Introducing Gateway.

Your agent needs to check a payment. You hand it a key that can also refund it.

It needs to draft a page. That key can archive every page.
Not anymore.

Gateway gives every agent one key with only the tools you grant. MCP servers and your own APIs, same key. Change the grant when the job changes. The agent never sees your real credentials.

Beta customers already run over 95,000 calls each through it.

Watch it work ↓

Your browser does not support the video tag.

🔗 View on Twitter

💬19🔄20❤️181👀34445📊73

⚡ Powered by xgo.ing

Get your key: https://t.co/ISjg4Wcgc5 Read our blog: https://t.co/jhhOA0o3GH

mem0 (@mem0ai) — Tue, 15 Sep 2026 15:21:15 GMT

Get your key: mem0.ai/gateway
Read our blog: mem0.ai/blog/give-ever…

💬0🔄0❤️3👀96

⚡ Powered by xgo.ing

We’re launching Gateway. Your agent needs to check a payment. Its API key can also refund it. It ne...

mem0 (@mem0ai) — Tue, 15 Sep 2026 15:15:17 GMT

We’re launching Gateway.

Your agent needs to check a payment. Its API key can also refund it.
It needs to draft a page. The same key can let it delete the workspace.

Access to a tool shouldn’t mean permission to do everything it allows.

Gateway lets you grant your agent only the actions you intend.

Check the payment. Draft the page. Keep the rest off-limits.

See it in action ↓

Your browser does not support the video tag.

🔗 View on Twitter

💬2🔄5❤️14👀583📊9

⚡ Powered by xgo.ing

https://t.co/9neuXbZZkU

mem0 (@mem0ai) — Mon, 14 Sep 2026 18:18:37 GMT

x.com/i/article/2099…

💬4🔄8❤️157👀16159📊64

⚡ Powered by xgo.ing

Rule of thumb: Create a field only for something you’ll filter, validate, or aggregate. Keep the res...

mem0 (@mem0ai) — Thu, 10 Sep 2026 14:33:06 GMT

Rule of thumb: Create a field only for something you’ll filter, validate, or aggregate.
Keep the rest as text.

That’s the tradeoff: flexibility at write time vs. reliability at query time.

mem0.ai/blog/structure…

💬0🔄0❤️3👀247📊1

⚡ Powered by xgo.ing

Why this matters: In 12 default-flow tests, Mem0 saved a stable preference (“I prefer dark mode”) ev...

mem0 (@mem0ai) — Thu, 10 Sep 2026 14:33:05 GMT

Why this matters:
In 12 default-flow tests, Mem0 saved a stable preference (“I prefer dark mode”) every time.
But it saved a temporary work update (“the migration is blocked”) only 2/12 times.
With the same blocker written as structured fields, it saved 12/12 times.

💬1🔄0❤️4👀240📊2

⚡ Powered by xgo.ing

How you store agent memory determines what your product can reliably do. Unstructured memory is fas...

mem0 (@mem0ai) — Thu, 10 Sep 2026 14:33:01 GMT

How you store agent memory determines what your product can reliably do.

Unstructured memory is fast to capture.
Structured memory is easy for code to query.
The best agents use both, on purpose.

Your browser does not support the video tag.

🔗 View on Twitter

💬11🔄2❤️41👀2613📊21

⚡ Powered by xgo.ing

https://t.co/TcM1ujYofB

mem0 (@mem0ai) — Wed, 09 Sep 2026 16:37:01 GMT

x.com/i/article/2097…

💬15🔄40❤️438👀40820📊212

⚡ Powered by xgo.ing

Everyone's watching the model race, and it's close enough right now that a percentage point barely d...

mem0 (@mem0ai) — Tue, 01 Sep 2026 17:32:32 GMT

Everyone's watching the model race, and it's close enough right now that a percentage point barely decides who wins. But the agent you can actually leave alone beats the agent that's 2% smarter on a benchmark and forgets everything the second the session closes.

What's the dumbest thing your coding agent has re-asked you this month?

💬0🔄0❤️0👀194

⚡ Powered by xgo.ing

So, what did we learn from this? The caching example was one decision, corrected in one moment. But...

mem0 (@mem0ai) — Tue, 01 Sep 2026 17:32:31 GMT

So, what did we learn from this?

The caching example was one decision, corrected in one moment. But delegation means the agent isn't making one decision, it's making dozens decisions for a full length of the run.

An agent working async in a cloud VM, pulling from a task queue, isn't getting corrected mid-thought by someone sitting next to it. That's not a place for guessing. That's a place for certainty, because it remembers the last time it guessed wrong.

💬1🔄0❤️0👀207📊1

⚡ Powered by xgo.ing

Coding agents are having a moment for one boring reason: They finally ship working code without you ...

mem0 (@mem0ai) — Tue, 01 Sep 2026 17:32:28 GMT

Coding agents are having a moment for one boring reason: They finally ship working code without you babysitting every line.

This isn't autocomplete anymore. This is delegation and it only works if the person or system you're delegating to doesn't need the full backstory every single time.

💬4🔄1❤️22👀1467📊8

⚡ Powered by xgo.ing

https://t.co/oZ39c4XZ2r

mem0 (@mem0ai) — Thu, 27 Aug 2026 17:21:14 GMT

x.com/i/article/2093…

💬10🔄7❤️93👀10225📊44

⚡ Powered by xgo.ing

What happens when the people building the future of AI sit around the same table? We're bringing a ...

mem0 (@mem0ai) — Wed, 26 Aug 2026 20:31:31 GMT

What happens when the people building the future of AI sit around the same table? 
We're bringing a small group of tech leaders for an intimate evening of AI, agents, memory and the realities of building at the frontier. 
Just smart people, good food and cool conversations!
If this interests you, come join us. 
Request an invite -> luma.com/mem0-22d7

💬1🔄0❤️9👀1263📊3

⚡ Powered by xgo.ing

In Mem0, you can model all three using the same add() and search() primitives: user_id for personal...

mem0 (@mem0ai) — Wed, 26 Aug 2026 15:02:34 GMT

In Mem0, you can model all three using the same add() and search() primitives:

user_id for personal facts
run_id and outcome metadata for events
agent_id for shared agent routines

Learn more about types of memory in our latest blog: 👉mem0.ai/blog/semantic-…a

💬1🔄0❤️4👀200📊2

⚡ Powered by xgo.ing

Procedural memory: stores repeatable steps for how to do something. Say “a deployment process” or “...

mem0 (@mem0ai) — Wed, 26 Aug 2026 15:02:33 GMT

Procedural memory: stores repeatable steps for how to do something.

Say “a deployment process” or “the five steps required to add a new API endpoint”.

Its failure mode is different: the procedure outlives the system it was designed for and keeps running without revalidation, much like being tattooed on and never questioned again.

💬1🔄0❤️2👀224📊2

⚡ Powered by xgo.ing

If you've ever watched Christopher Nolan's Memento, you already understand this. Leonard can't form...

mem0 (@mem0ai) — Wed, 26 Aug 2026 15:02:31 GMT

If you've ever watched Christopher Nolan's Memento, you already understand this.

Leonard can't form new memories. So, he splits everything into three systems: permanent facts, records of what happened, and rules he follows without question.

Most AI agents don't make that distinction and treat memory as one flat bucket.

But memory has to answer three very different questions:

What's true?
What happened?
How do I do this?

If you answer all three the same way, the agent quietly gets worse over time. 🧵

💬3🔄4❤️39👀3495📊15

⚡ Powered by xgo.ing

hermes memory setup → pick mem0. Built-in MEMORY.md and USER.md keep running alongside it by default...

mem0 (@mem0ai) — Tue, 25 Aug 2026 15:47:23 GMT

hermes memory setup → pick mem0. Built-in MEMORY.md and USER.md keep running alongside it by default. Three modes: Mem0 Cloud, self-hosted server via Docker, or fully OSS in-process with your own LLM, embedder, and vector store.
Read more: docs.mem0.ai/integrations/h…

💬2🔄0❤️8👀1959📊6

⚡ Powered by xgo.ing

Your @NousResearch Hermes agent has a 2,200-character memory. Its notes live inside the instructio...

mem0 (@mem0ai) — Tue, 25 Aug 2026 15:45:49 GMT

Your @NousResearch Hermes agent has a 2,200-character memory. 

Its notes live inside the instructions it loads at session start, so the space never grows. Hit the cap and the next thing it learns gets rejected - to remember something new, it has to delete something it already knew.

Mem0 plugs in as a memory provider and runs alongside the built-in files. The notes stop competing for space.

Watch the meter hit 100% here -

Your browser does not support the video tag.

🔗 View on Twitter

💬15🔄16❤️242👀19043📊93

⚡ Powered by xgo.ing

Once you’ve found the configuration that works perfectly for your use case, you're done. Apply the ...

mem0 (@mem0ai) — Wed, 19 Aug 2026 15:19:18 GMT

Once you’ve found the configuration that works perfectly for your use case, you're done.

Apply the exact configuration directly to your project. Go from experimentation to production with absolute confidence.

Try the Playground out for yourself here: app.mem0.ai/dashboard/play…

💬0🔄0❤️3👀172📊1

⚡ Powered by xgo.ing

But why do you need to tune your settings? Because memory isn't one-size-fits-all. Tweak your add o...

mem0 (@mem0ai) — Wed, 19 Aug 2026 15:19:17 GMT

But why do you need to tune your settings? Because memory isn't one-size-fits-all.

Tweak your add or search config in real time: what to remember, how strict a match has to be, extract facts or store verbatim.
Dial it in until it actually matches what your agent needs.

💬1🔄0❤️3👀240📊2

⚡ Powered by xgo.ing

Still crossing fingers about how memory will actually behave in your app? Introducing the new Playgr...

mem0 (@mem0ai) — Wed, 19 Aug 2026 15:19:16 GMT

Still crossing fingers about how memory will actually behave in your app?
Introducing the new Playground by Mem0, which allows you to test, tune, and perfect your context window without writing a single line of code.

Your browser does not support the video tag.

🔗 View on Twitter

💬7🔄2❤️25👀3390📊14

⚡ Powered by xgo.ing

Once you’ve found the configuration that works perfectly for your use case, you're done. Apply the ...

mem0 (@mem0ai) — Tue, 18 Aug 2026 15:13:56 GMT

Once you’ve found the configuration that works perfectly for your use case, you're done.

Apply the exact configuration directly to your project. Go from experimentation to production with absolute confidence.

Try the Playground out for yourself here: app.mem0.ai/dashboard/play…

💬0🔄0❤️1👀118

⚡ Powered by xgo.ing

But why do you need to tune your settings? Because memory isn't one-size-fits-all. Tweak your add o...

mem0 (@mem0ai) — Tue, 18 Aug 2026 15:13:55 GMT

But why do you need to tune your settings? Because memory isn't one-size-fits-all.

Tweak your add or search config in real time: what to remember, how strict a match has to be, extract facts or store verbatim.
Dial it in until it actually matches what your agent needs.

💬0🔄0❤️2👀192

⚡ Powered by xgo.ing

Still crossing fingers about how memory will actually behave in your app? Introducing the new Playg...

mem0 (@mem0ai) — Tue, 18 Aug 2026 15:13:54 GMT

Still crossing fingers about how memory will actually behave in your app?

Introducing the new Playground by Mem0: app.mem0.ai/dashboard/play…, which allows you to test, tune, and perfect your context window without writing a single line of code.

💬3🔄1❤️12👀1028📊7

⚡ Powered by xgo.ing

Thanks for the deep dive into Mem0's architecture by @_alejandroao! 🧠

mem0 (@mem0ai) — Mon, 17 Aug 2026 13:57:51 GMT

Thanks for the deep dive into Mem0's architecture by @_alejandroao! 🧠

Alejandro AO 🤗@_alejandroao

I took Mem0 apart to understand how AI agent memory really works.

00 Introduction
55 What is agent memory
26 Persistent memory stores
06 LLM-based ingestion
30 Semantic + BM25 retrieval
19 Entity boosting
38 Local open models

A practical architecture deep dive 🧵

Your browser does not support the video tag.

🔗 View on Twitter

🔗 View Quoted Tweet

💬1🔄3❤️22👀3357📊5

⚡ Powered by xgo.ing

Context Corner 2.0 is here! 🍽️🤖 Loved the first conversation on context, memory, and agents? We’r...

mem0 (@mem0ai) — Fri, 14 Aug 2026 04:50:18 GMT

Context Corner 2.0 is here! 🍽️🤖

Loved the first conversation on context, memory, and agents? We’re back for round two, with even more time around the table and an incredible lunch to match!

If you're a developer building, experimenting, or stuck on the frontlines of AI memory and agents, come hungry and ready to chat.

📍 San Francisco, CA
🗓️ Friday, Aug 14 | 12:00 PM - 1:30 PM PDT

Here's the registration liluma.com/mem0-a82xo5aF

💬1🔄2❤️36👀3576📊10

⚡ Powered by xgo.ing

That is the real debugging surface. Memory should not be a giant prompt stuffed with old conversatio...

mem0 (@mem0ai) — Thu, 13 Aug 2026 15:31:41 GMT

That is the real debugging surface.
Memory should not be a giant prompt stuffed with old conversations.

It needs its own layer.
That’s what we’re building at Mem0.

💬0🔄0❤️0👀183

⚡ Powered by xgo.ing

Most failures people blame on “the model” actually happen somewhere in this pipeline. - Extraction ...

mem0 (@mem0ai) — Thu, 13 Aug 2026 15:31:40 GMT

Most failures people blame on “the model” actually happen somewhere in this pipeline.

- Extraction missed the fact.
- The system stored temporary noise.
- Update logic failed to resolve a conflict.
- Retrieval pulled the wrong memory.

💬1🔄0❤️0👀188📊1

⚡ Powered by xgo.ing

Dumbledore didn’t keep every memory rattling around in his head. He extracted the important ones int...

mem0 (@mem0ai) — Thu, 13 Aug 2026 15:31:36 GMT

Dumbledore didn’t keep every memory rattling around in his head. He extracted the important ones into a Pensieve, then pulled them out when needed.  

Roughly how a good AI memory layer works too. Want to know what happens to today’s message while recalling 3 sessions later?

Your browser does not support the video tag.

🔗 View on Twitter

💬3🔄5❤️32👀3129📊11

⚡ Powered by xgo.ing

Mem0 plugin is now listed in Cursor Marketplace: add it once, and every chat can recall decisions, f...

mem0 (@mem0ai) — Wed, 12 Aug 2026 15:19:37 GMT

Mem0 plugin is now listed in Cursor Marketplace: add it once, and every chat can recall decisions, fixes, and conventions from across your whole codebase history. Not just this session.

The plug-in is here: cursor.com/marketplace/me…

Your browser does not support the video tag.

🔗 View on Twitter

💬7🔄9❤️47👀3973📊19

⚡ Powered by xgo.ing

https://t.co/CuxhcNmG3w

mem0 (@mem0ai) — Thu, 06 Aug 2026 14:58:48 GMT

x.com/i/article/2085…

💬9🔄27❤️207👀14622📊93

⚡ Powered by xgo.ing

Supersede and Merge run on every plan, as memories are added. Synthesis is opt-in (Pro and above) an...

mem0 (@mem0ai) — Wed, 05 Aug 2026 15:30:02 GMT

Supersede and Merge run on every plan, as memories are added. Synthesis is opt-in (Pro and above) and runs on a schedule, so new patterns can take up to a day to show up.
Read the full blog here: 👉mem0.ai/blog/stale-ai-…F

💬0🔄0❤️2👀301

⚡ Powered by xgo.ing

The interesting part is Synthesis. "Grip endurance is the ceiling" and "the engine (ski and row) is ...

mem0 (@mem0ai) — Wed, 05 Aug 2026 15:30:01 GMT

The interesting part is Synthesis.
"Grip endurance is the ceiling" and "the engine (ski and row) is a strength" were never written in any single memory.
Dream distilled them from the pattern across sessions, so the coach could confidently prescribe a grip-focused block.

💬1🔄0❤️2👀335📊2

⚡ Powered by xgo.ing

We tested whether AI agent memory has to get messier over time, or if it can clean itself up. Here's...

mem0 (@mem0ai) — Wed, 05 Aug 2026 15:29:46 GMT

We tested whether AI agent memory has to get messier over time, or if it can clean itself up. Here's what happened when we ran Mem0's new Dream feature on a Hyrox athlete's training log.

Your browser does not support the video tag.

🔗 View on Twitter

💬7🔄7❤️42👀4293📊20

⚡ Powered by xgo.ing

Try it now: https://t.co/PzaIP7PvAH Read our blog: https://t.co/a5HaMhqjXF

mem0 (@mem0ai) — Tue, 04 Aug 2026 15:14:50 GMT

Try it now: app.mem0.ai/dashboard/dream
Read our blog: mem0.ai/blog/dream-bac…

💬0🔄0❤️7👀570📊1

⚡ Powered by xgo.ing

Most active projects carry a few memories that quietly contradict each other. We built something for...

mem0 (@mem0ai) — Tue, 04 Aug 2026 15:13:19 GMT

Most active projects carry a few memories that quietly contradict each other. We built something for that!

Introducing Dream, the background layer that keeps mem0 memory clean, current, and true.
 → Merge: facts that repeat get folded into one clear memory
→ Supersede: the old fact stays as history, the new fact becomes the truth
→ Synthesis: not one memory — a pattern built from many, traced back to each one

Excited to share that we're live today on Pro and Enterprise!

Your browser does not support the video tag.

🔗 View on Twitter

💬10🔄5❤️99👀10177📊40

⚡ Powered by xgo.ing

We cut vector search latency by 70x, from 8–14 seconds to ~110–120 ms. The biggest unlock? Moving ve...

mem0 (@mem0ai) — Mon, 03 Aug 2026 15:12:42 GMT

We cut vector search latency by 70x, from 8–14 seconds to ~110–120 ms. The biggest unlock? Moving vector search to @turbopuffer  and keeping Postgres focused on what it does best.

If you're building memory infra for AI agents, this one's  for you: mem0.ai/blog/how-we-cu…

💬9🔄13❤️104👀7702📊42