Meridian is configured entirely through environment variables. There are no config files with special formats or UI settings to hunt down — every option is a key-value pair in a .env file. The installer creates these files and populates them during the credential walkthrough, but you can edit them at any time.
Where configuration lives
Meridian uses two env files, one per process boundary:
| File | Used by |
|---|
~/.meridian/.env | The Rust daemon — core settings and PM credentials |
services/.env | The Python agents — LLM endpoint and Jira credentials |
OPENROUTER_API_KEY for cloud LLM fallback. In practice you never need to touch it directly — the installer keeps it in sync with the values you provide during setup.
Editing configuration
To open the daemon config in your $EDITOR, run:
This opens ~/.meridian/.env. To edit the Python agents config, open services/.env directly:
After saving, restart the relevant daemon for the change to take effect:
Environment variable reference
Core daemon settings
These variables control how the Rust daemon reads screenpipe data and classifies sessions. All of them are optional — the defaults work out of the box.
| Variable | Default | Description |
|---|
SCREENPIPE_DB | ~/.screenpipe/db.sqlite | Path to screenpipe’s SQLite database. Meridian opens this read-only. |
MERIDIAN_DB | ~/.meridian/meridian.db | Path where Meridian writes its own session database. Created on first run. |
POLL_INTERVAL_SECS | 60 | How often (in seconds) the daemon checks screenpipe for new frames. |
CLASSIFICATION_ENABLED | true | Set to false to run activity tracking and categorisation only, with no MLX server needed. |
MLX_SERVER_PORT | 7823 | The port the persistent MLX inference server listens on. |
CLASSIFICATION_TIMEOUT_S | 120 | Per-session inference timeout in seconds. Increase this if you see timeout errors on long sessions. |
CLASSIFIER_BACKEND | mlx | Classification backend. Set to mlx when using the MLX inference server (recommended for Apple Silicon). |
Setting CLASSIFICATION_ENABLED=false lets you run Meridian in activity-tracking mode. The daemon still reads screenpipe frames, builds app sessions, and assigns activity categories — it just skips the ticket-linking step and doesn’t require the MLX server to be running. This is useful if you’re not ready to configure PM credentials or just want to explore the session data first.
Jira
All three credential variables are required to enable the Jira connector. JIRA_PROJECT_KEYS is optional — omitting it syncs across all projects your account can access.
| Variable | Default | Description |
|---|
JIRA_BASE_URL | (none) | Your Atlassian instance URL, e.g. https://your-org.atlassian.net. |
JIRA_EMAIL | (none) | The email address associated with your Atlassian account. |
JIRA_API_TOKEN | (none) | An Atlassian API token. Generate one at id.atlassian.com/manage-profile/security/api-tokens. |
JIRA_PROJECT_KEYS | (all projects) | Comma-separated project keys to scope ticket lookups, e.g. KAN,ENG. Leave empty to search all accessible projects. |
GitHub
Both credential variables are required to enable the GitHub connector. GITHUB_REPOS is optional — omitting it covers all repositories in your org.
| Variable | Default | Description |
|---|
GITHUB_TOKEN | (none) | A GitHub personal access token with repo scope. |
GITHUB_ORG | (none) | Your GitHub organisation name. |
GITHUB_REPOS | (all org repos) | Comma-separated list of org/repo slugs to scope issue lookups, e.g. your-org/api,your-org/web. Leave empty to search all org repos. |
Linear
The API key is required to enable the Linear connector. LINEAR_TEAM_IDS is optional.
| Variable | Default | Description |
|---|
LINEAR_API_KEY | (none) | A Linear API key. Generate one in your Linear workspace under Settings → API. |
LINEAR_TEAM_IDS | (all teams) | Comma-separated team identifiers to scope issue lookups, e.g. TEAM1,TEAM2. Leave empty to search all teams your key can access. |
Minimal working configuration
The smallest configuration that enables activity tracking, session categorisation, and Jira ticket sync is:
# Required for task classification and Jira sync
JIRA_BASE_URL=https://your-org.atlassian.net
JIRA_EMAIL=you@your-org.com
JIRA_API_TOKEN=your-api-token
# Enable classification (default: true — shown here for clarity)
CLASSIFICATION_ENABLED=true
Re-running credential prompts
The installer skips any variable that already has a value in the relevant .env file. To re-prompt for a specific credential, delete that line from the file and re-run the installer:
# 1. Remove the line you want to re-set (e.g. JIRA_API_TOKEN)
meridian config edit
# 2. Re-run the installer — it will prompt only for missing values
./install.sh --skip-permissions
--skip-env:
You can also edit ~/.meridian/.env directly with any text editor and then run meridian restart. The installer’s prompt flow is a convenience, not a requirement.