Articles
Guides, comparisons and research
End-to-end builds, setups for the stacks people use, honest comparisons with other ways to build agents, and long-form writing on how Mayura works.
20 articles shown
Build a coding agent that writes and runs code in a sandbox
An agent writes code and tests, runs them in an isolated machine of its own and fixes what fails: Docker on your laptop, E2B in production, nothing allowed by default.
- sandboxes
- coding agents
- permissions
- budgets
- testing
8 min read
- guide
Build a research agent that reads the web in a real browser
An agent opens pages in a real browser, kept to the sites you allow, and answers with sources, while a person watches live: local Chrome first, a hosted browser in production.
7 min read
- guide
Build a voice agent you can talk to
Speech in, an agent run, speech out: transcribe with Deepgram, answer with a tool-using agent, speak with ElevenLabs, with every call priced and bounded before it is made.
7 min read
- research
Deny by default: what an agent may do, and who decides
Mayura's permission model: exact grants for models, tools, effects and capabilities, scope from verified identity, and how it bounds what a misled model can do.
13 min read
- integration guide
Give agents E2B sandboxes in production
Move agents from local Docker sandboxes to E2B microVMs: templates, network allow-lists, a sandbox per run, lifetimes, costs, errors and cleanup.
9 min read
- research
Hard budgets: reserving the worst case before every model call
How Mayura keeps agent runs under their cost limit: each call's maximum is reserved before it starts and settled after, on one ledger shared by agents and workflows.
13 min read
- comparison
Mayura vs LangGraph.js
LangGraph.js and Mayura compared for production TypeScript agents: graphs and checkpoints, versus grants, hard budgets and steps that never run twice.
9 min read
- comparison
Mayura vs the OpenAI Agents SDK
The OpenAI Agents SDK for TypeScript and Mayura compared: handoffs, sessions and built-in tracing, versus grants, hard budgets and durable steps that never run twice.
9 min read
- comparison
Mayura vs the Vercel AI SDK
The AI SDK and Mayura compared for production agents: a toolkit for models, streaming and chat UI, and a runtime that enforces grants, budgets and durable steps.
9 min read
- integration guide
Mayura with Clerk
Sign users in with Clerk and let them call your agents: verify Clerk session tokens, map permissions to capabilities, and keep runs and memory per user.
7 min read
- integration guide
Mayura with Neon
Keep Mayura's state in Neon PostgreSQL over pooled, direct or WebSocket connections, and accept Neon Auth tokens on your agent server.
7 min read
- integration guide
Mayura with Supabase
Keep Mayura's workflows and run records in Supabase PostgreSQL and let users signed in with Supabase Auth call your agents.
6 min read
- research
Porting a Node.js agent framework to Workers, Deno, Bun and Vercel Edge
What stopped Mayura from running outside Node.js, the portable helpers that replaced Node APIs, what stays Node-only, and how CI holds every runtime to Node's results.
12 min read
- integration guide
Run Mayura on Cloudflare Workers
Serve agents from a Cloudflare Worker with runs that finish inside their request, workflows advanced by a Cron Trigger, and storage in D1 or PostgreSQL.
8 min read
- integration guide
Trace Mayura agents in Langfuse
Send agent runs and durable workflow runs to Langfuse as traces: models, tokens, cost and tool calls, with no prompt or output ever leaving your process.
8 min read
- guide
Build a refund agent that pays only after a person approves
An AI agent triages refund requests, a durable workflow checks policy, and nothing is paid until an operator approves the exact payment. Built from the approval-workflow starter.
7 min read
- integration guide
Deploy Mayura on Vercel
Run agents and durable workflows on Vercel Functions with Neon PostgreSQL: request-bound runs, workflows advanced by cron, and what we measured on a real deployment.
Updated 4 min read
- comparison
Mayura vs Mastra
Two TypeScript frameworks for AI agents and workflows, compared: what each is designed for, how they differ on permissions, budgets and durability, and when to choose which.
Updated 6 min read
- integration guide
Mayura with Next.js
Serve Mayura's agent API from a Next.js route handler, stream runs into React components, and run durable workflows next to your app.
4 min read
- research
Never twice: side effects in durable agent workflows
How Mayura dispatches each workflow step at most once, surfaces the calls it can't be sure about, and recovers from crashes without paying, emailing or deleting twice.
10 min read