⭐️⭐️⭐️ Open-source, self-hosted alternative to Linear and Plane. Project management and issue tracking where teams and AI agents work side by side to plan and ship products.
TypeScript agenticagileai-agentsbug-reportingbug-trackerganttissue-trackerjirajira-alternativekanbanlinearlinear-alternativeopen-sourceproduct-managementproject-managementroadmapself-hostedsprint-planningteam-collaborationtrello-alternative
# Shared .env for local dev and the self-hosting stack. Copy: cp .env.example .env
# Dev works with the values below as-is, except the secrets marked "generate one".
# Self-hosting needs real values for the secrets and public URLs; DATABASE_URL and
# S3_ENDPOINT are set by the compose file. Coolify uses docker-compose.coolify.yml.
# ── Postgres (docker-compose brings up the DB) ──
POSTGRES_USER=itsaplan
POSTGRES_PASSWORD=itsaplan
POSTGRES_DB=itsaplan
# Host port for the local dev DB. Change if 5432 is taken.
POSTGRES_PORT=5432
# ── Database connection (api + drizzle) ──
# Local; inside compose api reaches postgres by service name (postgres:5432).
DATABASE_URL=postgres://itsaplan:itsaplan@localhost:5432/itsaplan
# ── Public origins (each set once; every service derives from these) ──
# API_URL — api URL: auth handler, email links, and what the browser calls. An origin
# (its own subdomain), or a path on the web's host (https://app.example.com/api).
# APP_URL — web origin: CORS/auth trusted origins, passkey, email links.
API_URL=http://localhost:3000
APP_URL=http://localhost:3001
# Public origin Git providers deliver repository webhooks to, for an instance whose
# API_URL is reachable only from a private network. It changes only the payload URL
# (<origin>/webhooks/git/<id>); unset, the payload URL uses API_URL.
# GIT_WEBHOOK_BASE_URL=https://hooks.example.com
# ── Legal document URLs ──
# Shown on the logged-out screens; Google OAuth sign-in requires the privacy policy
# to be reachable. Point them at this instance's own documents. Leave empty to hide
# the legal notice. Read by web at startup, so a change needs a restart, not a build.
# PRIVACY_URL=https://example.com/privacy
# TERMS_URL=https://example.com/terms
# ── better-auth ──
BETTER_AUTH_SECRET=change-me-please-generate-a-real-secret # openssl rand -base64 32
# Passkey (WebAuthn) relying party. Defaults to the APP_URL hostname.
# In prod set to the public frontend hostname (no scheme, no port).
# PASSKEY_RP_ID=app.example.com
# PASSKEY_RP_NAME=Its a Plan
# Cookie domain override. Defaults to the parent domain of APP_URL, or to no domain
# (a host-only cookie) when API_URL is on the web's host; set only for multi-label
# TLDs or deep subdomains.
# COOKIE_DOMAIN=.example.com
# Client IP for the sign-in rate limit (docs/self-hosting.md, "Client IP"). By default
# X-Forwarded-For, taken only when it holds one address. Name another header your proxy
# sets itself, or list the proxies to skip from the right of an X-Forwarded-For chain.
# CLIENT_IP_HEADER=x-real-ip
# TRUSTED_PROXIES=10.0.0.5,172.18.0.0/16
# ── API (Elysia) ──
API_PORT=3000
# ── Database backup taken before migrations ──
# The migrate step dumps the database whenever a release has migrations to apply,
# and stops when the dump fails. Dumps older than the retention window are removed on
# the next run.
# BACKUP_RETENTION_DAYS=30
# Where the dumps are written. Under the compose files the db-backups volume pins it
# to /backups, so it only applies to a migration run outside a container.
# BACKUP_DIR=/backups
# Upgrade without a dump — for an operator who backs up by other means.
# SKIP_PRE_MIGRATION_BACKUP=1
# ── Migrations of an extending build ──
# A folder of a build's own migrations, which the migrate step applies after the core's.
# BUILD_MIGRATIONS_DIR=
# ── S3 object store: RustFS (issue attachments; metadata stays in the DB) ──
# Leave S3_* unset to disable uploads. The keys double as RustFS root credentials;
# the secret must be >= 8 chars. Upload limits are DB settings (god mode), not env.
S3_ENDPOINT=http://localhost:9000 # inside compose: http://minio:9000
S3_BUCKET=planner-attachments
S3_ACCESS_KEY_ID=minioadmin
S3_SECRET_ACCESS_KEY=minioadmin
S3_REGION=us-east-1 # required by the SDK, ignored by RustFS
# Set to false for a hosted store that serves only virtual-host-style buckets
# (Railway buckets, Cloudflare R2). RustFS needs the default.
# S3_FORCE_PATH_STYLE=false
# ── Secret encryption (AI provider API keys at rest) ──
# 32-byte AES key derived from this via SHA-256. Changing it makes stored
# credentials undecryptable. openssl rand -base64 32
APP_ENCRYPTION_KEY=
# ── Outbound request policy (apps/api + apps/worker) ──
# Server-side fetches of a URL that came from a request — a webhook target, an
# attachment to import, a repository host — are refused when the URL resolves to a
# loopback, link-local, or private address. Name a host here to trust it anyway.
#
# The case this exists for is a self-hosted repository host: its address is
# configuration rather than content, and on a self-hosted instance it is normally
# private, so without this a Gitea/Forgejo/GitLab on your own network cannot be
# connected at all. Exact hostnames, comma-separated; no wildcards or CIDR ranges.
# https is still required and the resolved address is still pinned.
#
# A named host is trusted by every outbound path, not only by the repository
# connection. The same list is read wherever a URL is vetted, so naming a host also
# makes it a permitted webhook target — validated by the api when a webhook is
# created or updated, and DELIVERED to by the worker — and a permitted source for
# attachment import. Anyone who can create a webhook or import an attachment can
# therefore reach a host you name here. Name only hosts you would let them reach.
#
# Set it on both the api and the worker: the api vets and the worker delivers, so a
# list on one and not the other silently applies to half the paths above.
# SSRF_ALLOWED_HOSTS=git.internal.example.com
# ── Telegram bot (apps/bot) — the bot token is stored in the DB, not here ──
# The service reads it with DATABASE_URL and APP_ENCRYPTION_KEY, set above.
# BOT_CONFIG_POLL_INTERVAL_MS=30000 # how often to re-read the bot settings
# BOT_CONFIG_RETRY_INTERVAL_MS=5000 # retry delay after the database was unreachable
# ── Background worker (apps/worker): webhooks, agent schedules, notifications ──
# Notification delivery reads the stored provider credentials, so the worker needs
# DATABASE_URL and APP_ENCRYPTION_KEY, set above.
# WEBHOOK_POLL_INTERVAL_MS=2000 # how often to poll for due deliveries
# WEBHOOK_BATCH_SIZE=20 # deliveries claimed/sent per tick
# WEBHOOK_TIMEOUT_MS=10000 # per-delivery HTTP timeout
# WEBHOOK_MAX_ATTEMPTS=8 # attempts before a delivery is dead-lettered
# WEBHOOK_DISABLE_THRESHOLD=20 # consecutive failures before a webhook is auto-disabled
# WEBHOOK_LEASE_SECONDS=120 # claim lease; a crashed delivery is retried after this
# WEBHOOK_CLEANUP_DAYS=30 # succeeded deliveries older than this are deleted
# WEBHOOK_CLEANUP_EVERY_TICKS=300 # run the cleanup once every N ticks
# NOTIFICATION_BATCH_SIZE=20 # notification deliveries claimed per tick
# NOTIFICATION_MAX_ATTEMPTS=5 # attempts before a delivery is marked failed
# NOTIFICATION_LEASE_SECONDS=120 # claim lease for crash recovery
# IMPORT_POLL_INTERVAL_MS=3000 # how often to poll for a due import job
# ── Background jobs (apps/api): the agent run queue, the auto-archive and share-expiry sweeps ──
# AGENT_RUN_POLL_INTERVAL_MS=2000 # how often to poll for due agent runs
# AGENT_RUN_TIMEOUT_MS=240000 # max run time until one is saved in god mode
# AGENT_RUN_MAX_ATTEMPTS=3 # attempts before a run is marked failed
# AGENT_RUN_LEASE_SECONDS=300 # claim lease for crash recovery; raised past the max run time
# AUTO_ARCHIVE_INTERVAL_MS=3600000 # how often the auto-archive sweep runs (own loop)
# AGENT_TRACE_PRUNE_INTERVAL_MS=3600000 # how often traces past their retention are deleted (own loop)
# SHARE_EXPIRY_INTERVAL_MS=60000 # how often expired share links are revoked (own loop)
# ── Chat with an external agent (apps/api): the feed its runner drains ──
# AGENT_CHAT_LEASE_SECONDS=300 # claim lease; an abandoned answer is served again after this
# AGENT_CHAT_MAX_ATTEMPTS=3 # attempts before an answer is marked failed
# AGENT_CHAT_CLAIM_WAIT_MS=25000 # claim wait, capped at 30s so the runner's 35s request timeout can finish
# AGENT_CHAT_CLAIM_POLL_MS=500 # how often that claim looks for work while it waits
# AGENT_CHAT_STREAM_POLL_MS=300 # how often an open stream looks for new events
# AGENT_CHAT_HISTORY_MESSAGES=20 # thread messages sent to the runner as context
# ── External agents with an invocation URL (apps/api): the api sends their tasks to it ──
# AGENT_INVOCATION_POLL_INTERVAL_MS=1000 # how often to look for tasks to send
# AGENT_INVOCATION_TIMEOUT_MS=15000 # wait for the endpoint's answer. With no answer, the task stays leased
# From address for outbound email. Required for Resend (verified sender); for SMTP
# defaults to the SMTP username. NOTIFICATIONS_EMAIL_FROM is read as a fallback.
# EMAIL_FROM=Its a Plan <[email protected]>
# ── Anonymous telemetry (apps/worker) — see TELEMETRY.md ──
# TELEMETRY_DISABLED=1 # opt out (any value but 0)
# DO_NOT_TRACK=1 # opt out, cross-tool convention
# TELEMETRY_DEBUG=1 # log the exact body before each send
# ── Web (Next.js) ──
# web reads API_URL, PRIVACY_URL and TERMS_URL from its environment at startup, so
# one image serves any instance. Local dev reads them from the separate apps/web/.env,
# which Next loads on its own.
# WEB_PORT=3001 # host port for the web app; change if 3001 is taken
# ── Published images ──
# The release the four services run. Unset means the newest one; pin it to stay on a
# known version. `docker compose up -d --build` ignores this and builds the checkout.
# VERSION=0.10.1