Self-Host Multica on Kubernetes

Multica is an open-source workspace for coordinating AI agents and their work: tasks, the agents running them, and the Slack conversations and GitHub pull requests attached to that work in one place. Self-hosting it keeps your agent instructions, credentials, and work history in your cluster. Edka installs the Multica chart and wires the database, secrets, and public hostname, on a cluster in your own cloud account.

What Multica Needs to Run

  • PostgreSQL 17 or 18 with the pgvector extension. Multica stores workspaces, tasks, agent runs, and embeddings here. Edka checks the version, the extension, and that the selected user owns the selected database before installing.
  • A public hostname on an Envoy Gateway traffic class. One hostname serves the web application, the API, WebSocket connections, and the Multica CLI.
  • Persistent storage for locally uploaded files, 10Gi by default.

Edka points Multica at an existing PostgreSQL installation from the Databases workspace and resolves the connection credentials from protected secrets, rather than running another database copy.

Resource Requirements

DeploymentCPU requestMemory requestLimits
Backend100m256MiOptional
Frontend100m256MiOptional

Both deployments run without CPU or memory limits by default, so they stay Burstable and can absorb spikes. Set explicit limits in the Resources tab when you want a ceiling. Budget separately for PostgreSQL. See the Hetzner instance types table.

What Edka Provisions

  • Multica backend and frontend. Separate deployments for the API/WebSocket server and the web application, with independent resource settings.
  • PostgreSQL connection. An existing installation, database, and user selected from the Databases workspace, with the connection string built from the user’s protected secret.
  • Stable runtime secrets. The JWT signing secret and the Slack, GitHub, and Git credential-encryption keys are generated during install and kept stable across updates.
  • Uploads volume. A persistent volume for locally uploaded files.
  • Gateway access. A hostname through an Envoy Gateway traffic class with Edka-managed TLS, carrying web, API, WebSocket, and CLI traffic.
  • Observability. Prometheus metrics for HTTP traffic, agent task throughput and failures, and LLM token usage and estimated cost, plus endpoints, runtime status, and logs in the Edka dashboard.

Integrations

  • Slack. Workspace admins connect one Slack Socket Mode app per Multica agent, so agents can be addressed from Slack channels. Edka generates and preserves the key that encrypts the stored Slack credentials.
  • GitHub. A GitHub App connection links pull requests to tasks and moves merged work to Done. Optionally enable CI status and mergeability on PR cards, which additionally needs the App ID and private key. Edka shows the exact setup and webhook URLs to configure on the GitHub App after installation.
  • Self-hosted Git. Optional Forgejo, Gitea, and GitLab connections for repositories outside GitHub.

Configuration Options

  • Database. Choose the PostgreSQL installation, database, and user. Edka validates version, ownership, and the pgvector extension at install time.
  • Authentication. Allow or close sign-ups, restrict registration to specific email addresses or domains, disable workspace creation after setup, and optionally enable Google OAuth.
  • Email. Set the sender address and an optional Resend API key for transactional email.
  • Resources and placement. Requests and optional limits per deployment, node pool placement, and automatic taint tolerations.
  • Auto-updates. Let Edka apply newer Multica chart versions using a patch, minor, major, or custom policy.

Deploy Multica

  1. Prepare a PostgreSQL 17 or 18 database with the pgvector extension and a dedicated owner user in the Databases workspace.
  2. Make sure the cluster has an Envoy Gateway traffic class and a DNS hostname for Multica.
  3. Select Deploy Multica, choose the database, and configure authentication.
  4. Deploy, open Multica, and create the first workspace.
  5. Enable the Slack or GitHub integration and connect your agents.

Read the Multica setup and operations guide for integration setup, secret handling, metrics, and troubleshooting. The upstream self-hosting guide and Multica repository cover the wider Multica platform.


Deploy Multica
icon related to Multica

Multica

Automation

Free

Category

AI Agents

Run Multica on your own Kubernetes cluster with PostgreSQL and pgvector, Slack and GitHub integrations, and per-agent metrics. Dependencies and setup explained.

Get started