Self-hosted

Your CRM. Your server. Your data.

Relaticle is open source under the AGPL-3.0 license. Run it on your own hardware with unlimited users and no per-seat cost, forever.

Terminal

# Download the compose file

curl -o compose.yml https://raw.githubusercontent.com/relaticle/relaticle/main/compose.yml

# Add the two secrets and your public URL

cat > .env << EOF

APP_KEY=base64:$(openssl rand -base64 32)

DB_PASSWORD=$(openssl rand -hex 24)

APP_URL=https://crm.example.com

EOF

# Start Relaticle

docker compose up -d

Requires Docker 20.10+ and Docker Compose v2. Environment variables, reverse proxy configs, and a manual (non-Docker) install are covered in the guide below.

Why run it yourself

The same reasons people self-host any serious open-source tool.

Your data stays yours

Every company, contact, opportunity, task, and uploaded file lives on infrastructure you control. Nothing is copied anywhere unless you choose to.

Source you can read and change

Relaticle is licensed AGPL-3.0. The full application is on GitHub, not a stripped-down edition, so you can audit it, modify it, and redeploy your fork.

One flat cost: your server

Unlimited users and unlimited records on every self-hosted install. There is no per-seat pricing and no user-count paywall to hit.

Three steps to a running CRM

Distilled from the full guide, which also covers Dokploy, Coolify, Traefik, and a Docker-free manual install.

  1. 1

    Get the compose file

    Download the published compose.yml, set the two secrets it requires (APP_KEY and DB_PASSWORD), and point APP_URL at the address you will serve from. Invitation and password-reset links are signed against that host, so they break if it is wrong.

  2. 2

    Start the stack

    Run docker compose up -d and Docker pulls five containers: the app, a queue worker, a scheduler, PostgreSQL, and Redis. Database migrations run automatically on startup.

  3. 3

    Create your admin account

    Run docker compose exec app php artisan make:filament-user, choose the app panel, and sign in with the account you just created.

AI, self-hosted

Your keys, your models

Rela is part of the self-hosted image, but it ships with no provider configured. You decide what it calls, if anything.

Bring your own provider key

Set ANTHROPIC_API_KEY or OPENAI_API_KEY and the assistant calls Claude or GPT with your own account, at your own usage rates.

Run it locally with Ollama

Point OLLAMA_BASE_URL at your Ollama server and set OLLAMA_MODEL to the tag you want in the picker. From inside the containers, that usually means http://host.docker.internal:11434, not localhost.

Either way, the model catalog in a hosted Relaticle Cloud workspace is provisioned by Relaticle. A self-hosted install starts with none of that: nothing answers until you configure a provider yourself.

Self-hosted or Cloud: how to choose

Both run the identical open-source Relaticle codebase. The real differences are who operates the server and who provisions the AI models.

  • Infrastructure

    Self-Hosted: Your own server, under your own operational control

    Cloud: Relaticle-managed infrastructure

  • Updates and backups

    Self-Hosted: You pull new images and back up PostgreSQL yourself (the guide has the pg_dump commands)

    Cloud: Handled by Relaticle, no maintenance window to schedule

  • AI models

    Self-Hosted: Bring your own provider key, or run a local model with Ollama

    Cloud: Cloud-hosted models ready immediately, no provider key required

  • Getting help

    Self-Hosted: Community support on Discord and GitHub issues

    Cloud: Email support

Self-hosting questions, answered

AGPL-3.0. The full application, the CRM, the AI assistant, and the MCP server, is on GitHub under that license: read it, audit it, modify it, and redeploy your own fork.
Pull the new images and restart the stack: docker compose pull then docker compose up -d. Database migrations run automatically on startup, so check the release notes for breaking changes before you upgrade.
A provider it can call. Set an API key for Claude or GPT, or point it at a local Ollama server. With neither configured, the assistant has no model to answer with.
On the PostgreSQL, Redis, and storage volumes defined in compose.yml, all running on your own server. Nothing is sent to Relaticle-operated infrastructure unless you choose the hosted Cloud plan instead.
The published minimum is 2 GB of RAM and 1 core, with 4 GB recommended for real workloads. Docker 20.10+ and Docker Compose v2 are required.
Yes. Both run the identical open-source codebase against the same schema, and every record type has a built-in CSV export, so moving between them is an export and re-import, not a proprietary migration.

Run it your way

Deploy the open-source image on your own server, or skip the ops and let Relaticle run it for you.