Which workflows are worth a team of agents.
Every post starts from a workflow: what goes in, which agent handles which part, and where the approach stops paying for itself.
One workflow, start to finish.
A scenario walked through end to end, with a runnable example you can open alongside the article.
Support Tickets: A Fixed Pipeline and a Routed Agent Team
Keep the high-volume path on a typed classify → draft → QA graph, let escalations get a coordinator-built team, and put the refund behind a confirmation gate the runtime enforces per call.
Competitive Monitoring: Isolate Each Source Before Comparing Them
One agent per source, structured claims out of each, and an aggregator that reads validated records instead of prose — so contradictions between a vendor post, a community thread, and a news article survive to the report.
Incident Postmortems: Parallel Investigation, Serial Judgment
Logs, deployments, and blast radius are independent evidence streams. A runTasks() DAG starts them together, keeps each result unmerged in taskResults, and hands the analyst validated structured payloads instead of prose.
Translation Drift: Route the Back-Translation to Another Model
Translate with one provider, back-translate with another through a model-routing rule, and let a reviewer return structured drift findings — then measure the reviewer itself with a versioned EvalSet.
From Transcript to Typed Action Items: Three Parallel Agents in TypeScript
Most meeting summarizers cram summary, action items, and sentiment into one LLM prompt. Here's a cleaner TypeScript shape: three specialist agents run in parallel, two return typed Zod output, and an aggregator merges them into one report.
Give Your TypeScript AI Agents Long-Term Memory with TencentDB-Agent-Memory
Wiring open-multi-agent's MemoryStore to TencentDB-Agent-Memory through its Hermes Gateway: a measured cross-run memory loop, and two upstream gotchas that decide whether anything gets stored.
Adding Multi-Agent Orchestration to a Vercel AI SDK App
Add multi-agent orchestration to an existing Vercel AI SDK app: the AI SDK streams tokens and talks to models while open-multi-agent's runTeam() decomposes the goal and coordinates the agents — sharing a single Next.js API route.
What it looks like once it runs.
What the tokens cost, where it slowed down, what got in the way — and the boundaries that only surface after enough runs.
Five Seams That Decide Whether a Workflow Needs a Team
Across OMA's runnable recipes the same five boundaries keep reappearing — and each one turns out to have a specific runtime mechanism behind it, from structured dependency payloads to an execution routing decision you can read back.
A 100% Local Multi-Agent Team in TypeScript (Ollama + Gemma, $0 API Cost)
Run a multi-agent team fully on your laptop in TypeScript — even the coordinator is a 5B Gemma over Ollama, at $0 API cost. With a real per-agent ledger, honest friction, and a hybrid cloud+local variant.
5 walls multi-agent frameworks hit: receipts from Mastra's year of .network() to Supervisor migration
Mastra spent a year migrating .network() to the Supervisor pattern. 5 engineering walls hit (context, routing, observability, nesting, performance), 18 GitHub issues as receipts. What it means for TypeScript multi-agent framework development.
How to Run a Mixed-Model AI Agent Team in TypeScript?
A practical walkthrough from a single-model team baseline to a mixed-provider production setup with live cost and latency monitoring, using open-multi-agent, the TypeScript-ecosystem answer to CrewAI.
Which framework fits which workflow.
Comparisons that cite their sources and say who each option suits, including the workflows where open-multi-agent is the wrong tool.
Best TypeScript Multi-Agent Frameworks in 2026: Choose by Workflow
A source-backed guide to six TypeScript options for multi-agent systems, matched to explicit graphs, agent UIs, handoffs, routed networks, all-in-one apps, and goal-driven task DAGs.
Multi-Agent AI in a TypeScript Service, Without the Python Sidecar
Three agents behind one Express route, in the Node process you already deploy: per-agent model tiers, validated JSON handoffs, a run-level token ceiling, and cancellation that actually cancels.
Goal-Driven Agent Orchestration vs Explicit Graphs: A TypeScript Framework Taxonomy
Most multi-agent framework reviews compare features. This post compares a different axis: where the framework places the decomposition cost. Goal-first frameworks pay it at runtime in tokens; graph-first frameworks pay it at design time in code.
How the runtime works, and what changed.
Mechanism walkthroughs and what shipped in each version.
Open Multi-Agent v1.14: Repair the Plan, Keep the Record
ReleaseOpen Multi-Agent v1.13: Route, Govern, Schedule, and Prove the Run
ReleaseOpen Multi-Agent v1.12.1: Evaluation, Offline Inspection, and a No-Key First Run
EngineeringGoal In, DAG Out: How Open-Multi-Agent Turns a Goal into a Task DAG
Taking this to production?
open-multi-agent is MIT-licensed and free to run yourself. When you need it delivered, integrated, or supported on a deadline, 元定义科技 (YuanASI) offers commercial delivery and support.