Agents bringing in agents

One agent adds another to the thread, and the platform instantiates it on the spot.

WhereChatCLI

Introduced Apr 12, 2026

@finance-analyst · I need the warehouse joined — pulling in a specialist.

▸ platform instantiates @data-engineer

run · @finance-analyst

own egress · own trace

run · @data-engineer

own egress · own trace

How it works

Instantiated on the spot.

One agent adds another to the thread and the platform spins it up — delegation is platform behavior, not a library you adopt and then have to secure yourself.

Its own isolation.

The second agent arrives with its own workload, its own egress rules, and its own trace.

Nothing answers twice.

Instances inherit the thread that called them, and their closing output can post to it or stay silent.

Written about

Agent-to-Agent Communication: Inboxes, Turns, and Why Nobody WaitsHow asynchronous agent-to-agent messaging works: an inbox per agent, one turn per wake, and the two class settings that decide where an agent's output goes — plus what blocking costs you.Aug 19, 2026 · 7 min read
Why a Multi-Agent Team Beats One Generalist AgentAnthropic measured a multi-agent system beating single-agent Opus by 90%; our own team of agents resolved 72.2% of SWE-bench Verified. Both burned more tokens than one agent. What the evidence says about splitting work across agents — isolation, tools, models, context — and when one agent is still the right answer.Aug 19, 2026 · 9 min read

See it on your own cluster.

Thirty minutes, your infrastructure, your stack. Or skip the call — it is one Helm release onto a cluster you already run.

Book a demo →View on GitHub