Build AI agents visually with Langflow or Flowise
Short answer: use a visual agent builder. You connect components on a canvas, test the result in a chat playground, and publish it as an API. Langflow and Flowise are two open-source builders you can host on Railway from a template. Both come with Postgres and Redis, and both let you build agents and document chatbots without writing a backend.
Deploy Langflow on RailwayWhen a visual builder helps
If you build apps with AI tools, the agent logic often ends up as scattered prompt code. A visual builder keeps it in one place. The templates list these uses:
- prototype and ship agents with tools such as web search, APIs and code, without writing a backend,
- retrieval-augmented chat over your own documents, exposed as an API,
- embeddable chat widgets and prediction APIs for your own apps,
- publishing flows as MCP servers for Claude Desktop, Cursor and other MCP clients (Langflow).
You bring your own model provider keys. The builder does not include model access.
Which AI app builder templates people deploy
| Template | Creator | Deployments | Services |
|---|---|---|---|
| flowise-railway | Henry Heng's Projects | 8,512 | 1 |
| Langflow | – | 2,852 | 2 |
| flowise-with-n8n-ai-automation | derekcheungsa's Projects | 1,974 | 3 |
| AnythingLLM | Timothy Carambat's Projects | 1,972 | 1 |
| Dify | Jack | 1,871 | 11 |
| Langflow | alphasec | 1,284 | 2 |
| FlowiseAI | Protemplate | 782 | 2 |
| Langflow Pre-releases | – | 358 | 2 |
Railway marketplace data as of 2026-11-19. Full ranking: Best AI app builder templates on Railway.
Langflow or Flowise
| Langflow | Flowise | |
|---|---|---|
| Pinned version | 1.12.5 | 3.1.4 |
| Services | Langflow, Postgres, Redis | Flowise, Flowise Worker, Postgres, Redis, Bucket |
| How it runs flows | 2 Gunicorn workers in one service; Redis shares the job queue | Queue mode: the main server queues runs in Redis and workers execute them |
| Files | Volume at /app/langflow |
Railway Bucket |
| First login | admin with LANGFLOW_SUPERUSER_PASSWORD |
First person to register becomes owner |
| Scaling | Raise LANGFLOW_WORKERS and resources; replicas are not possible while the volume is attached |
Raise Worker replicas or WORKER_CONCURRENCY; keep one main server |
| Licence | MIT | Apache-2.0, enterprise folder separate |
Deploy Langflow, step by step
- Deploy the template. It creates Langflow, Postgres and Redis, with generated admin credentials and a stable encryption key.
- Wait for the first deploy to pass its healthcheck. Migrations and starter projects can take a few minutes.
- Sign in. Open the public URL as
adminwithLANGFLOW_SUPERUSER_PASSWORDfrom the service’s Variables tab. - Add provider keys. Go to Settings, then Global Variables. Do not change
LANGFLOW_SECRET_KEYafterwards, or stored credentials stop decrypting. - Build a flow. Start from a starter project, connect components, test it in the Playground, then serve it as an API or MCP tool.
- Embed it. To use the chat widget on a website, add the site’s origin to
LANGFLOW_CORS_ORIGINS. For a custom domain, also updateLANGFLOW_MCP_BASE_URL.
For Flowise, open the public URL as soon as the deploy is healthy and create the owner account before sharing the link. Add credentials under Credentials, build a chatflow or Agentflow, and call its prediction API. The queue dashboard is at /admin/queues for logged-in users. To send invitation and password-reset e-mails, set the SMTP variables.
What persists
Langflow keeps flows, users, messages and encrypted variables in Postgres, and uploaded files, knowledge bases and logs in the config directory on a volume. Flowise keeps flows, users, credentials and executions in Postgres, and uploaded and generated files in a Railway Bucket. Both migrate the database on start, so back up Postgres before an upgrade and do not roll back to an older version against a migrated database.
Costs
Railway bills usage. The figures below come from the “Resource estimate” sections of each template’s spec. They are estimates, not measurements, and use Railway’s prices of $10 per GB of RAM per month, $20 per vCPU per month and $0.15 per GB of volume per month.
| Idle estimate | Typical estimate | Note | |
|---|---|---|---|
| Langflow (2 workers) | About 1.5 GB RAM, about $17 per month | About 2.4 GB RAM, about $32 per month | Each worker uses 0.6 to 1 GB RAM; Hobby is the minimum plan |
| Flowise (queue mode) | About 0.75 GB RAM, about $9 per month | About 1.65 GB RAM, about $25 per month | Three worker replicas: about $45 to $60 per month; Hobby is the minimum plan |
Langflow’s spec says the Trial’s 1 GB per service is below two workers, and that setting LANGFLOW_WORKERS=1 squeezes it into Trial for evaluation. Model usage is billed separately by your provider.
Limits and licence notes
- Langflow is MIT. Flowise is Apache-2.0, except
packages/server/src/enterprise, which has a commercial licence. Enterprise features need a key and stay off by default. - Flowise has no environment-based admin bootstrap, so register the owner first.
- Keep the Flowise main server at one replica, because it also runs the schedule beat. Scale workers instead.
- Redis 8 is available under RSALv2, SSPLv1 or AGPLv3; PostgreSQL uses the PostgreSQL License.
- These are community templates, not affiliated with the Langflow project or FlowiseAI, Inc.
Other options
See the AI app builder ranking for other builders, and the AI agents ranking for agent templates. For a workflow tool instead of an agent canvas, read the n8n queue mode guide. More guides are listed at /guides/.
FAQ
How do I log in to Langflow after deploying?
Open the public URL and sign in as admin with the password in the Langflow service variable LANGFLOW_SUPERUSER_PASSWORD. Add your model provider keys under Settings, then Global Variables. They are encrypted with LANGFLOW_SECRET_KEY, which you must never change after the first boot. New sign-ups stay inactive until an admin activates them.
How do I log in to Flowise? Is there a default admin?
There is no environment-based admin. The first person to register on the public URL becomes the owner, so create the owner account before you share the URL. Add model provider keys under Credentials. They are encrypted with FLOWISE_SECRETKEY_OVERWRITE, which you must never change after the first boot.
What do they cost to run on Railway?
Railway bills usage. The templates' specs contain estimates, not measurements, using Railway's prices of $10 per GB of RAM per month and $20 per vCPU per month. Langflow is estimated at about $17 per month idle and $32 typical, with Hobby as the minimum plan. Flowise in queue mode is estimated at about $9 idle and $25 typical, with Hobby as the minimum. Model usage is billed by your provider.
Which licences apply?
Langflow is MIT. Flowise is Apache-2.0, except the packages/server/src/enterprise directory, which has a commercial licence; enterprise features need a licence key and stay off by default. Redis 8 is available under RSALv2, SSPLv1 or AGPLv3, and PostgreSQL uses the PostgreSQL License. Check the terms for your use.
Are my flows kept when the service redeploys?
Yes. Langflow keeps flows, users and encrypted variables in Postgres, and uploaded files, knowledge bases and logs on a volume. Flowise keeps data in Postgres and uploaded files in a Railway Bucket. Back up Postgres before an upgrade, because both migrate the database on start.