Ward Peeters (@wardpeet) profile picture

OSS lead at @mastra. Web Performance and building stuff is my JAM. Herder of 🐐🐐🐐. (he/him)

Antwerp, Belgium
Joined May 2016
Photo by Ward Peeters (@wardpeet)Photo by Ward Peeters (@wardpeet)Photo by Ward Peeters (@wardpeet)Photo by Ward Peeters (@wardpeet)Photo by Ward Peeters (@wardpeet)Photo by Ward Peeters (@wardpeet)Photo by Ward Peeters (@wardpeet)Photo by Ward Peeters (@wardpeet)Photo by Ward Peeters (@wardpeet)Photo by Ward Peeters (@wardpeet)
Ward Peeters retweeted
i remember when 1000 was a lot
WE JUST HIT 3M DOWNLOADS 📈📈📈
4
1
39
996
Ward Peeters retweeted
Lets go @AamishBeg 👏 we use observational memory a lot at Mastra so we feel firsthand how well it works, but it's eye opening to see it in concrete numbers with detailed categorized breakdowns at different context lengths. Hell yea Aamish! Super exciting to see that even with top scores, there are clear areas to take it even further
Agent memory gets harder the longer a chat runs. We wondered: how well does Observational Memory hold up at 100K, 1M and 10M tokens? So we ran it on BEAM, the long-term memory benchmark, with default settings. Turns out it has the top published scores at 100K and 1M:
4
15
571
Welcome to the team 💪
I'm joining @mastra as a design engineer! super excited to work with this team, making our agent tooling look as good as it works :)
1
5
198
Knowledge is power!
Today we're announcing an early access program for Mastra Knowledge! At Mastra we build lots of internal Slack agents. But when we built our first company knowledge graph, it vastly improved agent quality. Grab time here to pilot with our team: calendly.com/mastra-team/mas…
85
Let's go!!
We’ve worked with hundreds of teams shipping Mastra agents into production. Now, it's your turn. We're introducing Mastra Launch and offering hands-on FDE support to get your agent live within 90 days. Let's chat: calendly.com/mastra-team/lau…
1
2
144
Sounds like a pretty good deal
Replying to @deepfates
use mastra. you get: - all the core agent primitives (agent loop, workflows, subagents, memory, processors) - harness primitives (subagents, dynamic agents, goals, crontab/scheduled, durability, signals, skills, sandboxes, browser and web use) - observability (traces, logs, evals, datasets, annotations) plus a studio to see all this in. and it’s an open source framework that works across infra/model providers
5
281
Ward Peeters retweeted
Replying to @lydiahallie
claude code finally starting to catch up to @mastra code
2
7
844
Ward Peeters retweeted
is Jev a revolutionary decision model or just an if-statement? @abhiaiyer added it to every primitive in Mastra to find out we cover Jev and all the other news in this week's Agents Hour
2
2
2
10
1,269
JEV JEV JEV JEV
In the 72 hours after Jev dropped, we shipped a workflows classifier, tool approval classifier checks, classifier as judge scorers, and a classifier as a processor. Thinking about a model router too. Come hang out and chat w/ us about it on Th: luma.com/mastra-dwi3
1
373
People think China's inference is slow, but DeepSeek v4.1 Flash is flying and dirt cheap. Frontier models should be fixing their limits... Astra/Fable fly through your usage; even Sol can get you limited pretty quickly nowadays.
2
221
We're in BETA! The factory gets better every day; customize it to your liking! Let's go!
Today we're launching Mastra Factory. When we started building a factory, we weren't sure what to expect. But over the last few weeks it's has been writing 30% of our PRs, closing 60% of issues. Factory is free + open-source, now you can try it too: producthunt.com/products/mas…
2
4
464
You can now scale Mastra on your own:
Today we're excited to launch the Mastra Helm chart Self-host Mastra with the same architecture we use at scale in our platform:
3
304
Ward Peeters retweeted
Today we're launching skill search for Mastra agents! Give your agents dozens of skills, let it search and load on-the-fly:
23
9
3
111
21,947
Probably the main reason to use Mastra. It just works!
Mastra is built to work with whatever stack you're already running. Check out all 80+ integrations at mastra.ai/integrations.
2
149
Ward Peeters retweeted
Multi-turn evals:
Today we're excited to launch multi-turn evals for Mastra agents! Now you can grade a whole conversation using gates and LLM-as-judge scorers:
1
4
24
2,787
Ward Peeters retweeted
Mastra is AG-UI compatible! 🎉 The full agentic feature set now runs on your @mastra agents, local and remote: - Self-healing A2UI - Live shared state - Human-in-the-loop interrupts - Background tasks and agent memory surfaced live - Voice, multimodality and more Bring your Mastra agent to ANY AG-UI frontend - including React, React Native, Angular, Slack, Teams and more. See more (repo, docs): go.copilotkit.ai/mastra-agui
5
17
1
84
5,387
Finally, I can torment my friends
We’re excited to launch iMessage support for Mastra agents! Now your agents get blue chat bubbles and triple dot typing indicators:
6
421
Ward Peeters retweeted
We put together a new guide on deploying Mastra to Kubernetes: • Share run state across pods with Redis and Postgres • Run durable agents so any pod can observe or resume the same run • Autoscale with a HorizontalPodAutoscaler
3
1
1
18
1,354