Multiplayer AI harnesses are having a moment.
We've been using one internally at @CoreSpeedHQ for about a month. The biggest change wasn't that we could run more agents. It was where our team started talking.
Before this, we had OpenClaw and Hermes in Slack. It seemed like the obvious setup: put agents where everyone already is.
But in practice, we still went back to Claude or Codex for serious work. Slack showed the conversation, but hid the actual work: the full session, repo context, branches, background tasks, dynamic workflows, and the progress of long-running jobs.
We were discussing in one place and steering in another.
So we reversed it. Instead of bringing agents into Slack, we brought the team into the agent workspace.
The harness we've been using is called Agent Ensemble. A month in, OpenClaw and Hermes in Slack are already gone. More of our team discussion now happens next to the work.
Slack is next.
How to switch from muse to Claude duet and lose nothing
We built @CoreSpeedHQ to keep not only the memory&context, but also your credentials, your logins, your keys, your tools, your devices access……everything
Never do a one-off work again and again
#duet#claudeduet#muse
another day, another personal agent lol.
this is why we’re building a shared foundation for all your agents at @CoreSpeedHQ. i want to try every new agent. i just don’t want to onboard each one like it’s their first day at the company.
We’re opening a research preview of Claude Duet, a new app where Claude works as a proactive agent.
It comes to you with what needs your attention, completes tasks you ask of it, keeps working when you step away, and checks with you before it acts on your behalf.
Everything your agents need, through one MCP. Now with the live web 🐟
@CoreSpeedHQ partnered with @Tiny_Fish so any agent can:
→ research many sites at once
→ work behind logins
→ watch pages and flag changes
CoreSpeed provides agent infra. TinyFish handles the web part.
this is the fun part of building CoreSpeed: how little you should have to build to make something like this work. the context, connections and permissions should already be there. you get to focus on what the agent should actually do.
Like this. I turned Consensus into an agent skill that works out of the box, now you don't need to buy 4 keys.
X, Reddit and Rednote access is built in through CoreSpeed MCP, and your own agent does the sorting.
One line to Install:
npx skills add corespeed-io/consensus
@yuaan1in@NuphosAI we built the same thing lol. this is how we kept our agent swarm running overnight on a 100k-line refactor PR, automatically switching accounts whenever one hit its session limit.
Think Costco for your agent’s SEO stack.
Introducing CoreSpeed MCP for SEO
Want your agent to handle SEO? Subscribing to these tools individually adds up fast:
1.Ahrefs Starter: $29/month
2.Semrush SEO: $139.95/month
3.Moz Starter: $49/month
4.Ubersuggest Individual: $29/month
5.Similarweb Competitive Intelligence: $199/month
That’s $445.95/month.🥹
CoreSpeed brings SEO data tools to your agent through one MCP connection, with pay-as-you-go pricing!
In our test, four tool calls cost about $0.04 in total.
One connection. One bill. Pay for what you use.🙌🏻
cloudflare just killed deno 🥲
“everything just works” is what i want from a runtime. typescript, type checking, linting, formatting, testing, explicit permissions. all built in, with sensible defaults.
i want to ship typescript apps faster with my agents, not waste time and tokens wiring up tools and configs for every new project.