Temporal

hosted-runtime

Properties

vendor: Temporal Technologies
url: https://temporal.io/
whatItIs: Durable execution platform that captures state at every step of a workflow so multi-step AI agents, MCP calls, and pipelines automatically resume from the point of failure instead of losing progress, available as self-hosted open source or Temporal Cloud.
hosting: both
pricingModel: open-source (MIT) self-host + usage-based Temporal Cloud ($1,000 free credits on signup)
modelProviders: agnostic
stateModel: durable workflow state / event history, replayed on resume
toolModel: multi-language SDKs; homepage lists agents, MCP, and AI pipeline orchestration as core use cases
includesBrowser: false
includesMemory: false
longRunning: unbounded durable workflows (automatic retry/resume across failures)
maturity: ga
launched: 2019
notes: Established durable-execution platform, not a September-2026 launch; founding year (2019, by Cadence co-creators Maxim Fateev and Samar Abbas) is corroborated by multiple investor/press profiles (Sequoia, Contrary Research) rather than stated directly on temporal.io itself. Surfaced as a baseline gap by its 2026-09-14 $550M Series E at a $12.55B valuation (source: temporal.io/blog/temporal-raises-usd550m-series-e-at-usd12-55b-valuation-ai), which the company frames explicitly around reliability for agentic AI (4,300+ paying customers including OpenAI, Snap, and NVIDIA; 1.9T billable actions processed in August 2026, up 350% YoY, per that announcement). Functionally analogous to already-tracked durable-execution runtimes Inngest AgentKit and Trigger.dev (whose own notes cite Temporal by name as a determinism comparator), but with a much larger install base, an MIT-licensed self-host path, and roots predating the current agent wave.

Properties

vendor

Temporal Technologies

url

https://temporal.io/

whatItIs

Durable execution platform that captures state at every step of a workflow so multi-step AI agents, MCP calls, and pipelines automatically resume from the point of failure instead of losing progress, available as self-hosted open source or Temporal Cloud.

hosting

both

pricingModel

open-source (MIT) self-host + usage-based Temporal Cloud ($1,000 free credits on signup)

modelProviders

agnostic

stateModel

durable workflow state / event history, replayed on resume

toolModel

multi-language SDKs; homepage lists agents, MCP, and AI pipeline orchestration as core use cases

includesBrowser

false

includesMemory

false

longRunning

unbounded durable workflows (automatic retry/resume across failures)

maturity

launched

2019

notes

Established durable-execution platform, not a September-2026 launch; founding year (2019, by Cadence co-creators Maxim Fateev and Samar Abbas) is corroborated by multiple investor/press profiles (Sequoia, Contrary Research) rather than stated directly on temporal.io itself. Surfaced as a baseline gap by its 2026-09-14 $550M Series E at a $12.55B valuation (source: temporal.io/blog/temporal-raises-usd550m-series-e-at-usd12-55b-valuation-ai), which the company frames explicitly around reliability for agentic AI (4,300+ paying customers including OpenAI, Snap, and NVIDIA; 1.9T billable actions processed in August 2026, up 350% YoY, per that announcement). Functionally analogous to already-tracked durable-execution runtimes Inngest AgentKit and Trigger.dev (whose own notes cite Temporal by name as a determinism comparator), but with a much larger install base, an MIT-licensed self-host path, and roots predating the current agent wave.