What is Dapr?
Dapr (Distributed Application Runtime) runs as a sidecar next to each service: a small process your code calls over HTTP or gRPC instead of talking to Redis, Kafka, a secret vault or another service directly. It offers a set of building blocks behind that one API: service-to-service calls with mutual TLS, publish and subscribe messaging, key-value state, secrets, actors and a durable workflow engine that resumes a run from its last finished step after a crash or a pod restart. The backing system is a configuration file, so swapping Redis for PostgreSQL does not change application code.
The project is Apache 2.0, written in Go and has been a CNCF graduated project since 30 October 2024. For agent builders it adds an alpha Conversation API for calling LLMs, documented durable-execution integrations for frameworks including the OpenAI Agents SDK, LangGraph, CrewAI, Pydantic AI and Google ADK, and Dapr Agents, a separate Python framework (v1.1.0) built on its workflows and actors.
Diagrid, a commercial company built around Dapr, sells Catalyst, a hosted Dapr service: a free tier, Cloud Plus at $31 a month billed annually, and dedicated or bring-your-own-cloud deployments from $1,199 and $1,599 a month.
How much does Dapr cost?
| Plan | Price | What you get |
|---|---|---|
| Dapr and Dapr Agents | Free | Apache 2.0 |
| Catalyst Cloud Plus | $31/month | Billed annually; 1M API requests/month |
| Catalyst Dedicated Cloud | From $1,199/month | Single-tenant, billed annually |
| Catalyst BYOC | From $1,599/month | Data plane in your cloud |
| Enterprise Server | Custom | Self-managed, air-gapped supported |
Runtime and Dapr Agents free under Apache 2.0. Diagrid Catalyst: Cloud free tier; Cloud Plus $31/month billed annually ($372/year; 1M API requests/month, 100 concurrent workflows per project); Dedicated Cloud from $1,199/month billed annually; BYOC from $1,599/month; Enterprise Server (self-managed, air-gapped) custom.
Is Dapr worth it?
Dapr answers a broader question than Temporal: not only whether a run resumes after a failure, but how dozens of services and agents call each other, share state and exchange messages the same way everywhere. That breadth is its strength for platform teams and its cost for everyone else. The runtime has about 26,200 GitHub stars, shipped v1.19.1 on 31 October 2026 and is a CNCF graduated project.
For agents, the useful parts are the workflow engine, pub/sub between agents, state for memory and identity-based access policies that say which agent may call which tool service. Dapr documents durable-execution integrations for the OpenAI Agents SDK, LangGraph, CrewAI, Pydantic AI, Google ADK, Strands Agents and Microsoft Agent Framework, so you can keep the framework you have. Dapr Agents, its own Python framework (v1.1.0 on 27 October 2026, Python 3.11 to 3.14), is much younger, with about 770 stars; the Conversation API it can use to call LLMs is still alpha.
The operational load is real. Even local development runs Redis, Zipkin, a placement service and a scheduler in Docker, and production means the Dapr control plane on Kubernetes plus a state store that supports workflows; the docs flag workflow complexity limits on Azure Cosmos DB and AWS DynamoDB. Dapr relies on the network boundary to protect its API by default, so API token authentication is opt-in. GitHub lists two high-severity advisories in 2026: an April path-traversal bypass of service-invocation access policies and an October operator flaw that disclosed secrets scoped to other apps.
Diagrid's Catalyst removes the hosting work. Cloud Plus at $31 a month billed annually covers 1M API requests and 100 concurrent workflows per project; dedicated single-tenant deployments start at $1,199 a month and are priced by concurrency, not execution volume. Temporal has the deeper workflow tooling and a larger agent integration list; Dapr is the better fit when the same sidecar also has to carry service calls, events and secrets across a polyglot estate.
Where it earns its place
A platform team at a retailer running 40 microservices on Kubernetes
Let new AI agents call existing order and inventory services safely
Agents run with a Dapr sidecar and call services by app ID over mTLS; access policies limit each agent to the operations it needs. The runtime is free; the effort is the control plane and policy files, and patching quickly given two high-severity advisories this year.
A Python team building a multi-agent triage system
Have agents hand work to each other through events without losing tasks on restart
Dapr Agents or an integrated framework such as CrewAI publishes tasks over pub/sub and runs each agent's steps as a durable workflow backed by Redis or PostgreSQL, so a crashed pod resumes from the last finished step.
An engineering manager at a 15-person SaaS startup
Get durable workflows without running a Kubernetes control plane
Diagrid Catalyst Cloud Plus at $31 a month billed annually covers 1M API requests and 100 concurrent workflows per project with a 99.9% uptime SLA; moving to a dedicated deployment starts at $1,199 a month.
Before you commit
- Setup
- Install the Dapr CLI and run dapr init, which needs Docker and starts Redis, Zipkin, placement and scheduler containers locally; dapr init -k installs the control plane on Kubernetes. Dapr Agents is pip install dapr-agents on Python 3.11 to 3.14.
- Integrations
- SDKs for .NET, Java, Python, Go, JavaScript/TypeScript and Rust; dozens of pluggable state stores, message brokers, bindings and secret stores; durable integrations for OpenAI Agents SDK, LangGraph, CrewAI, Pydantic AI, Google ADK, Strands Agents and Microsoft Agent Framework; MCP tool discovery in Dapr Agents.
- Data and privacy
- Self-hosted, state lives in the store you configure; workflow history can be signed with each sidecar's identity to detect tampering. The Dapr API relies on the network boundary unless you enable API tokens. Diagrid states SOC 2 Type II; BYOC keeps the data plane in your network.
- Cost at scale
- Self-hosting cost is your Kubernetes cluster and state store. Catalyst Dedicated Cloud on AWS lists $1,499 a month, or $1,199 billed annually, for up to 200 concurrent workflows with standard support; BYOC starts at $1,599 a month; on-premises Enterprise Server is quoted.
- Limits
- Workflows need a state store that supports them, with documented complexity limits on Cosmos DB and DynamoDB; the Conversation API is alpha. 442 open issues and pull requests on the runtime repository.
What’s good
- One sidecar API for workflows, messaging, state, secrets and service calls in any language
- mTLS and per-app access policies between services and agents
- Durable-execution integrations for OpenAI Agents SDK, LangGraph, CrewAI and others
- CNCF graduated, Apache 2.0, with a hosted option priced by concurrency
What’s not
- Heavy for one agent: sidecars, control plane and a workflow-capable state store
- Dapr API authentication is opt-in; the default trusts the network
- Two high-severity security advisories in 2026
- Conversation API is alpha and Dapr Agents is young
Best for
Platform teams on Kubernetes that want agents and ordinary services to share one model for durable workflows, messaging and secure service calls.
Dapr alternatives
Dapr: common questions
- Is Dapr worth it?
- Dapr pays off for a platform team already running many services on Kubernetes; for a single agent it is a lot of runtime to adopt for one workflow engine. We rate it 4.0 out of 5.
- How much does Dapr cost?
- Runtime and Dapr Agents free under Apache 2.0. Diagrid Catalyst: Cloud free tier; Cloud Plus $31/month billed annually ($372/year; 1M API requests/month, 100 concurrent workflows per project); Dedicated Cloud from $1,199/month billed annually; BYOC from $1,599/month; Enterprise Server (self-managed, air-gapped) custom.
- What are the downsides of Dapr?
- Heavy for one agent: sidecars, control plane and a workflow-capable state store; dapr API authentication is opt-in; the default trusts the network; two high-severity security advisories in 2026; conversation API is alpha and Dapr Agents is young.
- Who is Dapr best for?
- Platform teams on Kubernetes that want agents and ordinary services to share one model for durable workflows, messaging and secure service calls.
- How much does Dapr cost for a team or in production?
- Self-hosting cost is your Kubernetes cluster and state store. Catalyst Dedicated Cloud on AWS lists $1,499 a month, or $1,199 billed annually, for up to 200 concurrent workflows with standard support; BYOC starts at $1,599 a month; on-premises Enterprise Server is quoted.
- Is Dapr safe for company data?
- Self-hosted, state lives in the store you configure; workflow history can be signed with each sidecar's identity to detect tampering. The Dapr API relies on the network boundary unless you enable API tokens. Diagrid states SOC 2 Type II; BYOC keeps the data plane in your network.
- What are the best Dapr alternatives?
- The closest alternatives we have reviewed are Temporal, LangGraph and Microsoft Agent Framework.
Answers checked Nov 9, 2026.
Make Dapr? Show this rating on your site
