Dapr
dapr.io
- Free plan
- Windows
- macOS
- Linux
- API
- Self-hosted
What Dapr is
Dapr is an open-source runtime that provides APIs for distributed services, durable workflows and AI agents. Its workflows handle long-running stateful processes with retries, timers, human approvals and sagas; checkpointed plans help agents recover after failures. Service invocation offers HTTP or gRPC communication with discovery, tracing and error handling. Other APIs cover pub/sub, state, actors, jobs, secrets, configuration, bindings, distributed locks and cryptography. Dapr runs as a sidecar process or container, so application code can call its APIs without including runtime code. It supports Kubernetes and self-hosted deployments on Windows, Linux, macOS, and physical or virtual machines across cloud and edge settings. Security includes mTLS encryption in transit, access policies and certificate rotation. Stable SDKs are available for .NET, Python, Java, Go, PHP and JavaScript. Dapr itself is free and open source. Separate cloud offerings start at $1,199 a month for Dedicated Cloud and $1,599 a month for BYOC, each with the first 14 days free.
Who it suits
Dapr suits developers building distributed services, long-running workflows or AI agents who need a runtime for Kubernetes or self-hosted environments. Its stable SDKs cover six listed languages.
Good to know, and what to watch
- Workflows support retries, timers and human approvals.
- Service invocation supports HTTP and gRPC.
- Runs on Kubernetes or self-hosted machines.
- mTLS encrypts communication in transit.
- C++ and Rust SDKs are still in development.
- Dedicated Cloud starts at $1,199 a month.
- BYOC starts at $1,599 a month.
Plans
Every plan and priceEnterprise Server
Cloud Free
Dedicated Cloud
BYOC
Open source
Getting started
- Open the Dapr website
- Choose self-hosted deployment or a Kubernetes environment
- Select Windows, Linux, macOS, or Kubernetes hosting
- Use a stable SDK for .NET, Python, Java, Go, PHP, or JavaScript
- Run applications and inspect services with the Dapr CLI
Limits to check
The Cloud Free plan is for development and non-critical workloads. Dedicated Cloud and BYOC are paid plans, starting at $1,199 a month and $1,599 a month respectively.
The Club's review
Dapr is a sidecar runtime that gives distributed applications APIs for communication, state and durable execution. It is best for teams building services, workflows or AI agents across Kubernetes, cloud and edge infrastructure. Its breadth is a major advantage, while operating the runtime yourself requires engineering ownership.
Overview
Dapr places HTTP and gRPC APIs in a sidecar process or container, so application code does not embed the runtime. The component model separates those APIs from infrastructure resources: a team can switch implementations such as RabbitMQ or Kafka through component configuration instead of rewriting service code.
This is a broad platform, not a narrowly scoped workflow library. One runtime covers service invocation, publish-subscribe messaging, state, actors, jobs, secrets, configuration, bindings, distributed locks and cryptography. That consolidation can reduce bespoke integration work, but it also means adopting another system to configure, secure and monitor. Dapr originated as an open-source Microsoft project in 2019 and became a graduated CNCF project in October 2024.
Key features
Durable workflows and agents
Workflows support long-running, stateful processes with retries, timers, human approvals and sagas. Versioning and audit logs matter when a process must be changed or explained after it has started. Dapr Agents adds a Python framework for LLM-powered agents; checkpointed plans let agents recover after failures instead of losing their progress. This is a strong fit for business processes and agents that must survive crashes, restarts or outages, rather than for a simple request pipeline.
Service communication and resilience
Service invocation provides HTTP or gRPC calls with discovery, distributed tracing and error handling through one API. Configurable retries, back-offs, circuit breakers and timeouts give teams explicit failure controls instead of requiring every service to implement them independently. Pub/sub handles application-to-application events, while state, jobs and actors cover other common distributed patterns.
Security and observability
Security defaults include encrypted in-transit communication with mTLS. App IDs, namespaces, access policies, topic scoping, API tokens and automatic certificate rotation provide controls for isolating and governing services. Dapr emits metrics, logs and traces, with W3C Trace Context and OpenTelemetry support, so it can fit an existing observability approach. These defaults reduce the chance of an unsecured internal network, but teams still have to define policies and manage deployment configuration.
Developer tooling and integrations
Stable SDKs cover .NET, Python, Java, Go, PHP and JavaScript; C++ and Rust SDKs are still in development. The CLI starts applications with sidecars, shows sidecar logs and lists services in self-hosted and Kubernetes environments. Integrations include AWS, Azure, KEDA, GitHub Actions, the Dapr Kubernetes Operator, Kratix Marketplace and Argo CD. ASP.NET Core, Spring Boot, Flask and Express integrations lower the barrier for teams already using those frameworks.
Pricing
The open-source runtime is free, with a free plan for development and non-critical workloads. Cloud Free costs nothing and includes durable workflows, native integrations and workflow visualizers. It is the sensible starting point for evaluation, internal development and workloads where the team accepts self-management or free-plan constraints.
Dedicated Cloud starts at $1,199 a month, billed monthly, after a first 14 days free. It is single-tenant and fully managed, with compute, storage and networking included, 99.95% availability, and Select and Premium support. Choose it when operational responsibility and a dedicated environment justify a substantial recurring cost.
BYOC starts at $1,599 a month, billed monthly, also with a first 14 days free. The data plane stays in your cloud while the control plane runs in Diagrid cloud, adding data sovereignty and localization with 99.95% availability. It suits regulated or geographically constrained deployments, but costs more than Dedicated Cloud.
Enterprise Server uses custom pricing for your deployment. It supports on-premises, air-gapped and disconnected environments with namespace isolation, making it the relevant option when cloud hosting is not acceptable.
Diagrid offers enterprise support for open-source Dapr in three tiers, with commitments covering response times, security patches, architecture reviews and training. Those services are separate from the free runtime and are most relevant to organizations that need accountable operational assistance.
Platforms
Dapr runs self-hosted on Windows, Linux and macOS, on Kubernetes, and on clusters of physical or virtual machines. That range supports cloud and edge topologies, but portability also leaves deployment, upgrades and resource management to the operating team outside managed plans.
Who it's for
Dapr is aimed at platform and application teams standardizing service-to-service calls, messaging, state and resilient workflows across multiple languages or infrastructures. It is especially well suited to long-running approvals, sagas and AI agents that need checkpointed recovery. A small application needing one queue or a basic workflow may find the sidecar, component and policy model heavier than necessary.
Pros and cons
Pros
- Broad API surface: invocation, pub/sub, state, actors, jobs, secrets and more can share one runtime and component model.
- Failure-aware execution: durable workflows, retries, timers, circuit breakers and checkpointed agent plans address outages and restarts directly.
- Secure defaults: mTLS, certificate rotation and policy controls provide a solid baseline for service communication.
- Deployment flexibility: self-hosted machines, Kubernetes and managed cloud choices cover different infrastructure constraints.
Cons
- Operational responsibility: self-hosting means running sidecars, components, certificates and policies across environments.
- Managed pricing is high: Dedicated Cloud starts at $1,199 a month and BYOC at $1,599 a month.
- SDK coverage is uneven: C++ and Rust support is still in development, unlike the stable SDKs for six other languages.
- Platform breadth adds complexity: teams wanting only a focused workflow or messaging feature may be taking on more runtime than they need.
Alternatives
Temporal is a better fit when you want a workflow-focused platform with a usage-based managed plan: its Pay-As-You-Go plan is $50 a month with no base monthly fee and $150 in credits for 90 days.
Camunda suits teams comparing a process-orchestration product with a free self-managed development plan and production options.
Google Cloud Workflows is the more direct choice for Google Cloud users who prefer a managed service with an Always Free allowance of 5,000 internal and 2,000 external steps.
Bonita fits readers seeking a freemium process platform with a free tier capped at 150 process instances per month.
Joget is worth considering for low-user workflow applications, with a free tier for three users.
Flowable is an alternative open-source workflow and agent platform.
Bizagi Modeler is oriented to process modeling, with a free single-user plan and local storage.
Vercel is the alternative when the priority is web deployment rather than a distributed workflow runtime.
For a wider comparison, browse the Process Orchestration Software list.
Verdict
Choose Dapr if you need one runtime for secure service communication, stateful workflows and failure-resilient agents across Kubernetes or self-hosted infrastructure. Its main reason to look elsewhere is either operational complexity or the cost of managed cloud plans; a narrower workflow service can be simpler, and a basic open-source deployment still demands platform expertise.
The verdict
Dapr provides a broad set of APIs and deployment choices for services and durable workflows, with the runtime separated from application code. Teams considering the paid cloud offerings should weigh their listed starting prices against the free open-source option.
Questions people ask
Is Dapr free?
Dapr is free and open source. A Cloud Free plan is listed as free.
Where can Dapr run?
Dapr runs on Kubernetes and self-hosted Windows, Linux, or macOS machines, including physical and virtual machines.
Which SDK languages are stable?
Stable SDKs are available for .NET, Python, Java, Go, PHP, and JavaScript. C++ and Rust SDKs are in development.
What does the Cloud Free plan include?
It includes development and non-critical workloads, durable workflows, native integrations, and workflow visualizers.
What are the managed cloud prices?
Dedicated Cloud starts at $1,199 a month and BYOC starts at $1,599 a month. Both list a first 14 days free.
What the maker publishes
- AI agents
- Dapr enables AI agents to reason, call tools, and recover from failure with checkpointed plans. source
- APIs
- Dapr provides APIs for state management, secrets, bindings, actors, configuration, distributed locks, cryptography, jobs and conversations. source
- Building blocks
- Dapr provides APIs for workflow, service invocation, pub/sub, state, actors, jobs, secrets, configuration, bindings, distributed locks, and cryptography. source
- CLI
- The Dapr CLI runs applications with sidecars, reviews sidecar logs, and lists services in self-hosted and Kubernetes environments. source
- Component model
- Dapr components decouple APIs from underlying resources so implementations such as RabbitMQ and Kafka can be changed through component configuration without changing application code. source
- Deployment
- Dapr applications run on Kubernetes or self-hosted on virtual or physical machines across cloud and edge environments. source
- Founded
- 2019 source
- Frameworks
- Dapr has integrated frameworks including ASP.NET Core, Spring Boot, Flask and Express, and Dapr Agents is a Python framework for durable LLM-powered agents. source
- Hosting
- Dapr can run self-hosted on Windows, Linux or macOS and on Kubernetes or clusters of physical or virtual machines. source
- Integrations
- Official integrations include AWS, Azure, KEDA, GitHub Actions, the Dapr Kubernetes Operator, Kratix Marketplace, and Argo CD. source
- Microservices
- Service-to-service calls include service discovery, mTLS, retries, and distributed tracing through one API. source
- Observability
- Dapr emits metrics, logs and traces and supports W3C Trace Context and OpenTelemetry. source
- Open source origin
- Microsoft announced Dapr as an open source project on October 16, 2019. source
- Project status
- Dapr was accepted into CNCF incubation in 2021 and became a graduated CNCF project on October 30, 2024. source
- Pub/sub
- Dapr supports the publish-subscribe pattern between applications. source
- Purpose
- Dapr provides durable, verifiable execution so workflows and AI agents survive crashes, restarts, and outages. source
- Purpose
- Dapr is a portable runtime for building distributed applications across cloud and edge, combining event-driven architecture with workflow orchestration. source
- Resiliency
- Dapr supports configurable retries, back-offs, circuit breakers, and timeouts. source
- SDK languages
- Stable SDKs are available for .NET, Python, Java, Go, PHP, and JavaScript; C++ and Rust SDKs are in development. source
- Security
- Dapr security features are on by default and communication is automatically encrypted in transit with mTLS. source
- Security
- Dapr uses automatically enabled mutual TLS between sidecars, App IDs, namespace isolation and policy-based access controls for services, topics, components and secrets. source
- Security controls
- Dapr supports App IDs, namespaces, access policies, topic scoping, mTLS, API tokens, and automatic certificate rotation. source
- Service invocation
- Service invocation provides HTTP or gRPC communication with service discovery, distributed tracing and error handling. source
- Sidecar architecture
- Dapr exposes HTTP and gRPC APIs as a sidecar container or process, so application code does not include Dapr runtime code. source
Checked against the maker's pages on 1 Oct 2026.