Mastra
Mastra is an open-source TypeScript agent framework for durable workflows, memory, and built-in agent observability.
For TypeScript teams, Mastra is the most complete framework-level answer to agents that demo well and collapse in production — durable workflows, evals, and tracing arrive as real modules, and the October 2026 wave (Agent Controller GA, Mastra Connect, Trace Queries, Metrics API) closed gaps that used to need glue code. The friction is real: Python shops are out, and $250/mo Teams is a commitment before you've validated the idea. Start on the free Starter tier and move once your trace volume
Verified 7h ago · liveness 87/100 · cite: rightaichoice.com/tools/mastra
- TypeScript teams shipping multi-step agents that need durable execution, retries, and human approval
- Platform teams that want traces, evals, and scorers inside the framework instead of four separate services
- Regulated orgs needing RBAC, audit logs, VPC-isolated Postgres, and a flat-fee self-hosted license
- Developer platform teams exposing agents that other product teams build on
- Python-first teams unwilling to operate a TypeScript/Node runtime
- Single-turn chatbots where a direct model API call is cheaper and faster
- Throwaway prototypes that will never need durable execution, evals, or monitoring
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Mastra if your team is Python-first, or if a single-turn chatbot is the whole product — the workflow, memory, and eval machinery only pays off once agents genuinely run for days.
Observability events run past the plan's allotment at $10 per 100K on Starter ($8 per 100K on Teams), and long-running agents emit traces continuously
The $0 Apache 2.0 framework fits any TypeScript team at any size. The $0/mo Starter platform tier suits solo builders and early pilots; Teams at $250/mo fits funded teams with multiple squads that need SSO and SOC 2 docs. Enterprise is flat annual with no per-trace metering. Cheaper than building your own observability stack; comparable to assembling LangSmith or Braintrust alongside a framework, but with workflows included.
In short
Mastra — Mastra is an open-source TypeScript agent framework for durable workflows, memory, and built-in agent observability. Best for TypeScript teams shipping multi-step agents that need durable execution, retries, and human approval, Platform teams that want traces, evals, and scorers inside the framework instead of four separate services, Regulated orgs needing RBAC, audit logs, VPC-isolated Postgres, and a flat-fee self-hosted license. Free to start; paid plans from $250/mo.
What's new in Mastra
Checked 3 days agoAcross the latest 10 updates: 7 feature updates, 1 launch, 1 community discussion and 1 news mention.
Introducing Mastra Studio-specific SSO
Mastra Studio now supports SSO sign-in via WorkOS, Okta, Auth0, Clerk, or Google.
Introducing Trace Queries
Trace Queries surfaces slow or failed runs for coding agents to fix.
Introducing Trace Imports for Mastra Platform
Mastra Platform can now import and inspect existing trace history.
Introducing the Metrics API for Mastra Platform
Metrics API queries cost, latency, and errors by agent, thread, or time window.
Introducing Thread View for Mastra Observability
Thread View reads agent traces in the context of the conversation that produced them.
Introducing the Mastra Agency Program
Mastra launches an agency program of certified consultancies building Mastra agents, with five launch partners.
Multiplayer agents need fine-grained context
Mastra post explains fine-grained context (FCA) and how it applies to building multi-user agents.
Introducing Mastra Connect
Mastra Connect gives agents access to Linear, Notion, Slack, and 20 other services with a single key.
Agent Controller GA
Agent Controller is generally available, supporting long-running interactive agents with independent sessions and persistent threads.
Introducing VPC-isolated Postgres on Mastra Platform
Mastra Platform offers managed PostgreSQL for VPC-isolated deployments.
What people actually say about Mastra — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
81 mentions across 5 sources (Hacker News, YouTube, Product Hunt, GitHub, Lemmy) · researched Jul 30, 2026.
Average across the 5 sources that answered — each source counts once, not each post.
- +Built-in observability (evals, metrics, traces, logs) removes need for separate tools.
- +Typed agents with instructions, models, and tools defined in one file.
- +Graph-based workflow engine with .then(), .branch(), .parallel() methods.
- +Durable execution via Temporal for fault-tolerant long-running agents.
- +Human-in-the-loop workflow suspension and tool approval built in.
- −Severe supply chain attack compromised 140+ npm packages in June 2026.
- −Unpredictable behavior reported with the @mastra/ai-sdk beta.1 version.
- −TypeScript-only – no Python support limits adoption in ML teams.
- −Still young – 600+ open GitHub issues indicate ongoing rough edges.
- −Complex setup for durable execution requires Temporal infrastructure.
- • Temporal infrastructure costs for durable execution
- • Potential premium for cloud-hosted Mastra Studio
Viability Score
How well maintained and how widely used is Mastra? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Typed agents with id, name, instructions, model, and tools in one file
- Model router string format (provider/model) for thousands of models
- Durable workflows composed with .then() chaining and commit()
- Typed, retriable workflow steps with branches
- Human-in-the-loop suspension and tool approval inside workflows
- Harness / Agent Controller for long-running agents with independent sessions and persistent threads
- Memory with observational memory, semantic recall (topK), and thread-aware storage
- Signals subscribe agent threads to webhook or polling events
- Observability: traces, metrics, logs, evals, scorers, datasets, experiments
- Trace Queries to surface slow or failed runs for agents to fix
- Trace Imports and Thread View for inspecting existing trace history
- Metrics API for cost, latency, and errors by agent, thread, or time window
- Mastra Connect single-key access to Linear, Notion, Slack, and ~20 services
- Fine-grained context and authorization per user and per resource
- Local Studio and Mastra server running at localhost:4111
About Mastra
Mastra is an Apache 2.0 TypeScript framework for building AI agents and the production plumbing around them: workflows, memory, a server, and observability. You define an agent in one file — id, name, instructions, model string like openai/gpt-5.5, and tools — then chain multi-step work with createWorkflow().then().commit() so each step stays typed and retriable. It's aimed at engineering teams in TypeScript/Node who need agents that survive past the demo. The stack ships as opt-in modules: Agents, Workflows, Harness, Memory, Server, and Factory. Harness (Agent Controller, GA Oct 2026) coordinates multi-mode agents with shared state, threads, and storage — the same harness behind Mastra Code. Memory gives agents durable context via observational memory, semantic recall with topK retrieval, and thread-aware storage, and Signals (Oct 2026) wake, notify, and steer long-running agents from webhook or polling events. Factory pitches specialized agents that take software from issue to production. Observability is part of the framework rather than a separate tool: traces, metrics, logs, evals, scorers, datasets, and experiments live beside your agent code. The platform additions in October 2026 made that surface deeper — Trace Queries to surface slow or failed runs, Trace Imports for existing history, a Metrics API for cost, latency, and errors by agent or thread, and Thread View that reads traces in the context of the conversation. Mastra Connect opens Linear, Notion, Slack, and ~20 services with a single key, and fine-grained context controls what multi-user agents see. Positioning: the framework is free and self-hostable anywhere. The paid Mastra Platform meters observability events, CPU time, memory tokens, and data retention, while a licensed Enterprise self-hosted option keeps traces and prompts inside your VPC on a flat annual fee. Against LangGraph.js or hand-rolled orchestration, Mastra trades a smaller plugin ecosystem for one integrated mental model across
Behind the Verdict
Where Mastra earns its keep is the shipping phase. A team that already lives in TypeScript gets typed agents, durable workflows with retries and branches, and human-in-the-loop approval in the same repo as the tracing and evals — no stitching a workflow engine to a separate observability vendor. We'd pick it when an agent has to run for days, resume after a failure, and be debuggable by someone who didn't write it. October 2026 made the observability story notably stronger. Trace Queries surfaces slow or failed runs for coding agents to fix, Thread View reads a trace against the conversation that produced it, and the Metrics API cuts cost, latency, and error data by agent, thread, or time window. If your current stack is a dashboard you squint at after an incident, this is the upgrade. The pass cases are just as clear. Python-first teams should stay with the Python ecosystem rather than introduce a Node runtime for one agent. Single-turn chatbots and throwaway prototypes are cheaper with a direct model call. And if your hobby project generates heavy trace volume, the metered allowances — 100K events and 24 CPU hours on Starter — will push you toward Teams fast. On price: the Apache 2.0 framework costs nothing and self-hosts anywhere, so you can evaluate it without a contract. The Platform ladder is Starter at $0/mo (100K observability events, 24 CPU hours, 15-day retention), Teams at $250/mo (1M events, 250 CPU hours, 6-month retention, multiple teams, SSO, SOC 2 docs), and Enterprise on custom pricing with RBAC, audit logs, SLAs, VPC-isolated Postgres, and a flat annual self-hosted option. Closest alternative is LangGraph.js, which has a larger community of graph-pattern examples but leaves observability and eval tooling to you. Mastra's edge is coherence —
Researching Mastra? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Mastra actually fits — and what changes day-one when you adopt it.
Runs npm create mastra@latest with the openai provider, defines a weather-style agent and a createTool() tool, starts the dev server with npm run dev, and opens Studio at http://localhost:4111 to test the agent conversationally.
Outcome: A working typed agent with tool calling in a local Studio session the same day, with traces already visible.
Registers agents and workflows on a single Mastra server, adds fine-grained authorization per user and per resource, and turns on a persistent 24/7 server for internal teams to build against.
Outcome: Internal product teams consume versioned agents through one server with per-user scoping instead of each team standing up its own bot.
Wires in the Vitest integration, adds Experiment Tool Mocks so tests don't hit live APIs, and configures multi-turn evals with deterministic gates plus an LLM-as-judge scorer to assert whole conversations.
Outcome: Agent regressions get caught in CI before deployment, and experiments run repeatedly without burning real API spend.
Use Cases
- Build a customer-support agent with durable workflows that pause for human approval and resume
- Drop agent regression tests into CI with the Vitest integration before deploying
- Run multi-turn evals with deterministic gates plus LLM-as-judge scorers on existing TypeScript agents
- Deploy a Slack-integrated agent with channels and per-user authorization
- Use Factory to run issue-to-production workflows with Jira, GitLab, and incident.io follow-up
- Redact PHI and enforce access controls for HIPAA-regulated agent deployments
- Load tools and skills on demand with Tool Search and Skill Search to control token spend on long agent runs
- Expose agents to other product teams on a shared Mastra server with fine-grained authorization
Models Under the Hood
as of 2026-10-16
Limitations
- Mastra is an open-source TypeScript agent framework, so building agents with it targets JavaScript/TypeScript developers rather than Python users.
- The framework itself is free under an Apache 2.0 license (build and host agents anywhere), while the managed Mastra Platform is consumption-based: the Starter plan is $0/month with 100K observability events, 24 CPU hours, and 15-day retention, with overages charged (+$10/100K events, +$0.35/CPU hr); Teams is $250/month (1M events, 250 CPU hours, 6-month retention); Enterprise is custom pricing with RBAC, audit logs, and support/uptime SLAs, plus separately licensed self-hosted enterprise features at custom pricing.
- Self-hosted deployments retain full control of where models run through the framework's provider/model router.
as of 2026-09-30
Verification history
We have re-verified Mastra 23 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 23 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Mastra tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Starter
$0/mo
Ideal for
Solo builders and early pilots running a few agents through the managed platform without SSO needs
What this tier adds
Adds the hosted platform: 100K observability events, 24 CPU hours, 15-day retention, unlimited users and deployments
Teams
$250/mo
Ideal for
Funded teams with multiple squads that need SSO and SOC 2 documentation for procurement
What this tier adds
Raises quotas to 1M observability events and 250 CPU hours, extends retention to 6 months, and adds multiple teams, SSO, and SOC 2 docs
Enterprise
Custom
Ideal for
Regulated enterprises that need data to stay in their own VPC with flat annual pricing
What this tier adds
Custom volume and retention plus RBAC, audit logs, uptime SLAs, a dedicated support engineer, and a one-flat-annual-fee self-hosted license
Where the pricing makes sense
The company stage and team size where Mastra's pricing actually pencils out — and where peers do it cheaper.
The $0 Apache 2.0 framework fits any TypeScript team at any size. The $0/mo Starter platform tier suits solo builders and early pilots; Teams at $250/mo fits funded teams with multiple squads that need SSO and SOC 2 docs. Enterprise is flat annual with no per-trace metering. Cheaper than building your own observability stack; comparable to assembling LangSmith or Braintrust alongside a framework, but with workflows included.
Setup time & first value
How long it actually takes to get something useful out of Mastra — broken out by persona, not the marketing-page minute.
A new project with a working agent runs in minutes from one command (npm create mastra@latest). Expect under an hour to first tool call in Studio. Adding durable workflows with human-in-the-loop approval, memory configuration, or evals each take an afternoon of reading docs. Platform setup (deployment, SSO, authorization scopes) is a day or two for a platform engineer.
Switching to or from Mastra
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From a raw LLM SDK: move prompt strings into Agent instructions and replace ad-hoc chaining with workflow .then() steps for retry and durability
- →From LangChain.js: port chains to createWorkflow steps and .commit(), and reuse existing tool definitions via createTool()
- →From CrewAI: translate role definitions into specialized Mastra agents coordinated through Harness or Factory
- →From an in-house agent loop: keep your tools, wrap them with createTool() and zod schemas, and let Mastra handle memory and tracing
- →From cron-driven scripts: replace scheduled scripts with durable workflows that suspend and resume on human approval
- ↗To LangChain.js: rewrite workflows as chains and take your prompts and tools, trading built-in evals for a wider plugin ecosystem
- ↗To a direct provider SDK: keep simple agents and drop the framework if you never use durable execution or evals
- ↗To a Python framework: requires a full rewrite, since Mastra has no Python runtime
- ↗To an in-house stack: export traces and evals data through the platform before cancelling, since retention caps at 15 days on Starter
Integrations
Resources & Guides
- Documentationmastra.ai
Get started with Mastra
Full product docs from mastra.ai
- Learnmastra.ai
Build Your First AI Agent in TypeScript
Educational content from mastra.ai
- Resourcemastra.ai
Mastra Resources: Workshops, Books, and Research for AI Agents
Master the skills to build and launch AI agents with guided instruction, practical tools, and real-world examples.
- Guidemastra.ai
Overview
In-depth how-to from mastra.ai
- Resourcemastra.ai
Mastra Blog: AI Agents, Workflows, and TypeScript Engineering
Mastra.ai blog - Latest updates, tutorials, and insights on building AI agents.
- Resourcemastra.ai
AI Agent Builder Templates and Starter Kits
Spin up working examples of AI agents, workflows and MCP servers. Mastra Templates are starter kits for AI agent builders.
Tutorials & Learning
YouTube returned 6 videos for “Mastra”, and we withheld 6: 6 could not be judged, because “Mastra” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about Mastra.
Official links
Tools that pair well with Mastra
Common stack mates teams adopt alongside Mastra, with the specific reason each pairing earns its keep.
Microsoft Agent Framework
Microsoft Agent Framework is an open-source SDK for building agentic AI with workflows, checkpoints, and the GA Agent Harness runtime.
MetaGPT
Open-source multi-agent framework that assigns PM, architect, engineer and QA roles to GPTs for structured software tasks
OpenAI Agents SDK
A free, MIT-licensed Python framework from OpenAI for building multi-agent workflows with handoffs, guardrails, and tracing.
Featured Head-to-Head Comparisons
Mastra vs Vercel Ai Sdk
Guard Skills vs Mastra
Mastra is the right choice if you're building complex, production-grade AI agent systems with multi-step workflows, durable execution, and robust observability. Guard-skills is ideal if your primary concern is ensuring quality of AI-generated code, especially in WordPress/WooCommerce. They serve different needs and can even be complementary.
Mastra vs Value For Fable
If you need a full-stack agent framework with durable workflows, observability, and multi-agent orchestration, Mastra is the way to go — it's built for production. If you're on a tight budget and want to squeeze Opus-like reasoning from Sonnet with a structured prompting approach, Value-for-Fable gives you that at zero cost, but it's purely a prompting wrapper, not an agent framework. Choose based on whether you need infrastructure (Mastra) or cost optimization (VFF).
Alternatives to Mastra
View all
Microsoft Agent Framework
Microsoft Agent Framework is an open-source SDK for building agentic AI with workflows, checkpoints, and the GA Agent Harness runtime.
MetaGPT
Open-source multi-agent framework that assigns PM, architect, engineer and QA roles to GPTs for structured software tasks
OpenAI Agents SDK
A free, MIT-licensed Python framework from OpenAI for building multi-agent workflows with handoffs, guardrails, and tracing.
Frequently Asked Questions
Used Mastra? Help shape our editorial sentiment research.