Temporal
by Temporal Technologies
Open-source durable execution platform that keeps long-running AI agents and workflows alive through crashes, timeouts and failed tool calls
Temporal is a durable execution platform for engineering teams that run long-lived workflows and AI agents in production. Developers write workflows as normal code, and Temporal saves the state of every step so work resumes where it stopped after a crash or failed external call, which removes most hand-written retry and recovery logic.
Temporal is a durable execution platform: developers write workflows as ordinary code in one of eight SDK languages (Go, Java, .NET, PHP, Python, Ruby, Rust and TypeScript), and Temporal records the state of each step so a workflow resumes where it stopped after a server crash, a timeout or a failed call to an outside service. The core server is open source under the MIT licence and has about 24,000 GitHub stars. The company earns its revenue from Temporal Cloud, the managed service, which bills per million actions plus active and retained storage. Founders Samar Abbas and Maxim Fateev worked on Amazon Simple Workflow Service and Microsoft's Durable Task Framework, then built the open-source Cadence engine at Uber, before starting Temporal in 2019. The company now pitches the platform as the reliability layer for AI agents that run for hours, days or weeks and call many tools and models, and its docs include recipes for the OpenAI Agents SDK, Vercel's AI SDK, Google ADK, Strands Agents, MCP servers and Claude tool calling. On September 14, 2026 Temporal announced a $550 million Series E at a $12.55 billion valuation co-led by Lightspeed, seven months after a $300 million Series D at $5 billion. It reports more than 4,300 paying customers including OpenAI, NVIDIA, Netflix, Snap and JPMorgan Chase, 1.9 trillion billable actions in August 2026 and an annualized revenue run rate above $250 million. All of those figures are company-reported. Smaller rivals in durable execution include Inngest, Trigger.dev, Restate and Hatchet.
A platform or infrastructure engineering lead whose team is moving AI agents or multi-step business processes into production and keeps writing custom retry, queue and state-recovery code.
Workflows and agent runs survive process crashes and outside-service failures and resume from the last completed step, without the team building that recovery logic itself.
At a Glance
- Category
- AI Agents & Orchestration
- Pricing
- Usage-based, Free
- Target Market
- CTOs, Platform Engineering Teams, Enterprise Developers, AI Engineers
- Deployment
- Open-source, Self-hosted, Cloud-first
- Founded
- 2019
- Team Size
- 500+
- Customers
- 4,300+ paying customers (company-reported, September 2026)
Key Features
- ✓Durable execution
Persists workflow state at every step so execution resumes from the last completed step after a crash or outage, instead of restarting the whole job.
- ✓Built-in retries and failure handling
Declarative retry policies and timeouts on Activities replace the boilerplate error-handling code teams usually write around flaky APIs and LLM calls.
- ✓Signals, Queries and Updates
Lets running workflows receive outside input, such as a human approval, and expose their current state, which is how agent pauses and approvals are built.
- ✓Timers and Schedules
Workflows can sleep for days or run on a schedule without holding a process open, useful for reminders, follow-ups and recurring agent tasks.
- ✓AI agent framework recipes
Documented integrations for the OpenAI Agents SDK, Vercel AI SDK, Google ADK, Strands Agents, MCP and Claude tool calling make each LLM call a durable Activity.
- ✓Eight language SDKs
Go, Java, .NET, PHP, Python, Ruby, Rust and TypeScript SDKs let teams write workflows in the language their services already use.
- ✓Worker Versioning
Supports rolling out workflow code changes while older executions are still running, a common source of breakage in long-lived processes.
Capabilities
Use Cases
- •Durable AI agent loops
Each LLM call and tool call runs as an Activity, so an agent that fails on step 40 retries that step instead of rerunning the whole expensive chain.
- •Approval workflows for agents
An agent pauses for a human sign-off delivered through a Signal, waits as long as needed, then continues with full history recorded.
- •Distributed transaction orchestration
Payments, orders and supply-chain steps across microservices are coordinated with retries and compensation, so partial failures do not leave inconsistent records.
- •Data pipelines and ETL
Multi-stage ingestion and transformation jobs resume from the last good stage after an outage, cutting reruns and manual cleanup for data teams.
Ideal For
Best For
- ✓Long-running AI agents that call many models and tools and must not lose progress when one call fails
- ✓Human-in-the-loop agent flows that pause for approval using Signals and resume days later
- ✓Financial, order and supply-chain processes that coordinate transactions across several microservices
- ✓Data pipelines, scheduled jobs and infrastructure provisioning that need reliable retries and an execution history
Not Ideal For
- ✗Small teams or hobby projects: practitioners on Hacker News call it overkill for light use, and PeerSpot reviewers find Temporal Cloud costly for small organizations
- ✗Teams without senior engineers available to learn its determinism rules and workflow versioning, which reviewers name as the steepest part of adoption
- ✗Simple single-step background jobs, where a plain queue or a lighter tool such as Inngest or Trigger.dev needs less setup
Integrations
Deployment
Market & Ratings
4,300+ paying customers (company-reported, September 2026)
Market Analysis
Pros
- ✓Workflows resume from the last successful step after failures, and retry policies remove a lot of error-handling boilerplate (PeerSpot reviewers)
- ✓Scales to very high volumes: Snap moves 414 million Stories a day on it and OpenAI's usage grew 60-fold in under a year
- ✓Open-source MIT core with eight language SDKs and a responsive Slack community
- ✓Documented recipes for the major agent frameworks make it a practical reliability layer for AI agents
Cons
- ✗Steep learning curve; a PeerSpot reviewer said only senior developers navigate it effectively, and others call the documentation thin on examples
- ✗Workflow code must be deterministic for replay, and changing workflows can break compatibility with running executions
- ✗Self-hosting needs real infrastructure (Kubernetes plus Cassandra or Postgres), and Temporal Cloud can be costly for small organizations
- ✗The web UI is basic; reviewers want better workflow-state visualization, and one noted the images are not FIPS compliant
Pricing
Open source (self-hosted)
$0
- ✓MIT-licensed Temporal server
- ✓You run and pay for the infrastructure (Kubernetes plus Cassandra or Postgres)
Temporal Cloud Pay-As-You-Go
From $50 per million actions
- ✓No base monthly fee
- ✓$150 in credits for 90 days
- ✓SAML SSO included
- ✓Developer support
- ✓Storage billed separately
Volume discounts
From $25 per million actions
- ✓Self-service tiered pricing
- ✓Larger discounts with a commitment via sales
Business support add-on
From $500/mo or 10% of usage
- ✓2.5M actions included
- ✓2-hour P0 SLA in business hours
Enterprise / Mission Critical support
Contact for pricing
- ✓10M actions included
- ✓SCIM
- ✓30-minute 24/7 P0 SLA (15 minutes on Mission Critical)
- ✓Dedicated platform architect
The server is free and MIT-licensed; self-hosting costs are your own infrastructure, which PeerSpot reviewers say can be expensive. Temporal Cloud lists pay-as-you-go pricing from $50 per million actions (Workflow, Activity, Timer, Signal and other operations), dropping to $25 per million at volume, plus active storage from $0.042/GBhr and retained storage from $0.00105/GBhr. Temporal says no features sit behind plan upgrades; higher tiers buy support SLAs, included usage and SCIM. Prices checked on temporal.io/pricing on October 27, 2026.
Security & Compliance
Connect
Sources
This page was written from 8 sources, 6 on domains other than temporal.io.
- 1.temporal.io — temporal raises usd550m series e at usd12 55b valuation aivendor
- 2.temporal.io — pricingvendor
- 3.docs.temporal.io — ai cookbook
- 4.docs.temporal.io — develop
- 5.thenextweb.com — temporal 550m series e 12 55bn valuation lightspeed
- 6.origin.peerspot.com — temporal reviews
- 7.github.com — temporal
- 8.hn.algolia.com — search
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Notion Agents
Autonomous AI agents inside the Notion workspace that run on triggers and schedules for up to 24 hours, with a model router and credit-based pricing
Town
AI work assistants ("Townies") for every employee that handle email, calendar and routine work, and now hand tasks to each other across a team
Komodor Agentic Operations Platform
Agentic operations platform for SRE and platform teams: 50+ prebuilt Kubernetes, cloud and cost agents plus an SDK to run your own agents under shared governance
Stuut
AI agents that run accounts receivable and order-to-cash inside your existing ERP