A visual workflow builder for developing AI apps with diagrams instead of code. On a node-based editor, you chain prompts together, transform data along the way, and call external APIs — the whole flow is assembled by drag and drop. It supports major models including OpenAI, Claude, and LLaMA, and ships with integration nodes for more than 180 external services. Development, testing, and production can be managed as separate branches, and performance monitoring after a run is handled on the same screen. The provider is GBD Software as a Service Ltd, based in Hungary.
Key Features
- Node-based visual editor: Processing steps such as prompts, data transforms, conditional branches, and API calls are placed as “nodes” and connected with lines to form a workflow. Because the flow is visible as a diagram, it is easy to trace where things get stuck
- Prompt chaining: Instead of stuffing everything into one giant prompt, you split the work into small steps — summarize, classify, polish the wording — and chain them. Intermediate output can be passed as input to the next node, so accuracy can be tuned step by step
- Use multiple LLMs selectively: Supports major models such as OpenAI, Claude, and LLaMA. A different model can be assigned per node within a workflow, allowing patterns like “a cheap model for the draft, a high-performance model for the finish”
- More than 180 external API integrations: Nodes are provided for connecting to spreadsheets, chat tools, CRMs, databases, and other external services. This suits use cases where AI output is fed directly into business systems
- Branch management for development, testing, and production: Environments can be separated using the same idea as software development. You can try a new prompt without stopping production, then roll it out once it checks out
- Performance monitoring: Execution results, processing time, and error occurrences are recorded and can be reviewed later. This gives you material for comparing results before and after a prompt change
- Security designed for team use: GDPR compliant and supporting two-factor authentication (2FA). Per-member permission management is also included, built with collaborative work by multiple people in mind
Pricing
| Plan | Monthly price | Details |
|---|---|---|
| Free plan | $0 | A free tier for trying out the features |
| Paid plans | Contact for pricing | Expanded usage, team features, and more |
Pricing is as of August 2026. The official pricing page is generated dynamically, and specific figures could not be confirmed at the time of writing. Please check the official site for the latest pricing.
Pros & Cons
✅ Pros
- AI processing flows can be assembled without writing code, making it easier for non-engineers to take part in the design
- Because the flow is visible as a diagram, it is easy to pinpoint the node responsible when something goes wrong
- A model can be chosen per node, so the balance between accuracy and cost can be tuned in detail
- Development, testing, and production can be managed separately, supporting an approach that assumes real production use
- A free plan is available, so you can get a feel for it first
⚠️ Cons
- No concrete price list is published on the official site, so an inquiry is required before serious adoption
- Japanese-language information is still scarce, and the official documentation and help center are mainly in English
- As nodes pile up the diagram grows complex, and large workflows can become hard to follow
- Model usage fees are billed separately, so you need to estimate the actual running cost yourself
- As with visual tools in general, there are situations where writing code would be faster when fine-grained control is required
Comparison with Similar Services
| Item | AgentRunner | Dify | Flowise | n8n |
|---|---|---|---|---|
| Delivery model | Cloud | Cloud / self-hosted | Mainly self-hosted (OSS) | Cloud / self-hosted (OSS) |
| Main use | Building AI workflows and turning them into apps | LLM app development and RAG | Prototyping LLM apps | General business automation (AI is one part) |
| Editing method | Node-based visual | Visual plus configuration screens | Node-based visual | Node-based visual |
| Environment separation | Branch management for dev / test / production | Version management available | Limited | Production and test can be separated |
| External integrations | More than 180 API integrations | Tool integrations available | LangChain-family integrations | A very large number of integrations |
Who Is It For
- Anyone who wants to build an internal tool using AI but has no room to write code from scratch
- People who want to share the trial-and-error of prompt tuning across a team and improve it while keeping a history
- People who want to compare multiple LLMs and pick the best model for each step
- Teams that need environment separation and monitoring for real production use, not just a prototype
- Anyone who wants to wire AI output into existing business systems
Summary
AgentRunner is a workflow builder aimed at assembling AI processing as a diagram and carrying it all the way through to production. Prompt chaining, per-node model selection, more than 180 external integrations, and branch management for environments are gathered on a single screen, which shortens the distance from prototype to operation. On the other hand, there is no explicit price list on the official site and Japanese-language information is limited, so a realistic approach is to build one small workflow on the free plan first and confirm whether it fits your use case.