Rasa hosting

A conversational AI framework for assistants that must behave predictably.

Support & CXApache-2.0 (open source edition)from $18.99/mo

In short

Rasa hosting at AIHostBox means a dedicated, single-tenant Rasa instance on our infrastructure: we install it, pin it to a reviewed release, put it behind TLS on your domain, back it up nightly and keep it patched. Plans start at $18.99 per month and the admin account, the data and any API keys are yours.

Intents & entitiesDialogue policiesCustom actionsTestable flows

Rasa Open Source is the Apache-2.0 conversational AI framework from Rasa Technologies. You define intents and entities with training examples, write stories and rules for how conversations should go, and connect custom actions in Python for anything that touches your own systems. A trained model classifies each message and a dialogue policy chooses the next step. The result is an assistant whose behaviour you can test before it reaches customers, which is why banks, insurers and telecoms have built on it for years.

Two facts shape how to host it today. First, Rasa Open Source is in maintenance mode on the 3.6 line; new development, including the LLM-based CALM approach, happens in Rasa Pro under a commercial licence. Second, it is server software: rasa run exposes the REST API and channel webhooks on port 5005, a separate action server runs your Python code on 5055, and a tracker store keeps every conversation. Training is CPU-intensive and customers write at all hours, so neither belongs on a laptop.

Rasa hosting plans

ProRecommended minimum

8 GB RAM · 4 vCPU · NVMe

$18.99/mo

Order

MaxLarge document sets

12 GB RAM · 6 vCPU · NVMe

$27.99/mo

Order

Every plan is a dedicated single-tenant instance with NVMe storage, nightly off-box backups, TLS and managed patching. Quarterly, semi-annual and annual terms are discounted at checkout. Model tokens are not included — you bring your own key.

How we run Rasa

Two containers from the Rasa Open Source images — the Rasa server (rasa run) on port 5005 and the rasa-sdk action server on 5055 — with Postgres as the SQL tracker store.

What we configure for you

tracker_store in endpoints.yml set to type: SQL on the bundled Postgres, so conversations survive restarts
action_endpoint pointed at the action server on port 5055
--enable-api with --auth-token or --jwt-secret when you need the HTTP API
Channels in credentials.yml (REST, Slack, Telegram and others) with your tokens
--cors origins for a website chat widget
Project files and trained models on persistent storage, reachable over SFTP

Which plan fits

Pro (8 GB, 4 vCPU) runs the server, action server and tracker store with room to train a typical NLU and dialogue model. Max (12 GB, 6 vCPU) suits larger training sets and busier assistants. Training time grows with your data, so large projects may prefer to train in CI and upload the model.

The honest limits

Rasa Open Source is in maintenance mode and does not include CALM or other Rasa Pro features. Pipelines with large transformer components train slowly on CPU.

What people run Rasa for

Regulated support flows

Conversations that must follow an approved script.

High-volume deflection

Handle the top twenty questions without a model call each time.

Website and messaging channels

Connect one assistant to a web widget, Slack, Telegram or a plain REST client.

Backend-driven answers

Custom actions look up orders, balances or bookings in your own systems.

Licence and version policy

Item	Position
Upstream licence	Apache-2.0 (open source edition)
Sold as managed hosting	Yes
Tenancy	Single tenant — one customer per instance
Patching	Critical advisories within 24 hours, high severity within 72

Our full position on licences, including the applications we decline to sell, is on the licensing page.

Reading on Rasa

ComparisonSelf-hosted Intercom alternatives

Works well with Rasa

More Support & CX we host

Questions

Rasa hosting: your questions

Is Rasa Open Source still maintained?

It is in maintenance mode: the repository sits on the 3.6 line, and Rasa directs new projects to Rasa Pro and its CALM engine. We host the Apache-2.0 open-source edition, which remains sound for intent-based assistants. Rasa Pro needs a licence from Rasa itself, which we do not supply.

Where are conversations stored?

In a tracker store. The default keeps them in memory and loses them on restart, so we configure the SQL tracker store on Postgres. Rasa also supports Redis, MongoDB and DynamoDB; if you add Redis as a lock store, upstream warns it must not be the same Redis instance as the tracker store.

Can I train models on the instance?

Yes. Upload your project over SFTP and run rasa train there, or train in CI and upload the model archive to the models directory; rasa run loads the newest model on start. Training competes with live traffic for CPU, so run it outside busy hours on a production assistant.

How is the Rasa HTTP API secured?

The HTTP API is off unless the server starts with --enable-api. When you need it, we add --auth-token or --jwt-secret and keep the server behind our TLS proxy, as Rasa’s own docs recommend instead of exposing it directly. Channel webhooks continue to work on their own routes.

Can I move an existing Rasa project across?

Yes, if it runs on Rasa Open Source 3.x. Send us the project — config.yml, domain, data, actions and endpoints.yml — and, if you want history, a dump of your tracker store database. Projects still on Rasa 2.x need migrating to 3.x first, and Rasa Pro projects need their own licence.

Run Rasa without running a server

A dedicated Rasa instance, deployed and maintained by us. Month to month, cancel whenever.

Browse the catalogue Talk to a human