Integrations

Catalyst integrates with external services via CLI tools and MCP servers. Each integration requires credentials in your secrets config.

GitHub

Pull request creation, code review, and repository management via the gh CLI.

Setup: brew install gh && gh auth login

No Catalyst-specific configuration needed — gh uses its own authentication.

Related skills: /catalyst-dev:create-pr, /catalyst-dev:describe-pr, /catalyst-dev:merge-pr, /catalyst-dev:commit, /catalyst-pm:sync-prs

PR-Linear Sync

The /catalyst-pm:sync-prs skill correlates GitHub PRs with Linear issues — matching via branch names and descriptions, identifying orphaned PRs and issues, and flagging stale PRs.

Worktree Integration

Worktrees created with /catalyst-dev:create-worktree automatically set up branches with ticket references (e.g., PROJ-123-feature-name).

Linear

Ticket management and automatic status progression via the Linearis CLI.

Setup: npm install -g linearis + add apiToken and teamKey to secrets config.

Why CLI instead of MCP? Linearis uses ~1K tokens vs Linear MCP’s ~13K — a 13x reduction in context cost.

Related skills: /catalyst-dev:linear, /catalyst-pm:analyze-cycle, /catalyst-pm:analyze-milestone, /catalyst-pm:groom-backlog, /catalyst-pm:sync-prs

Automatic Status Updates

Workflow skills automatically update Linear ticket status as you progress:

Skill	Linear State
/catalyst-dev:research-codebase	In Progress
/catalyst-dev:create-plan	In Progress
/catalyst-dev:implement-plan	In Progress
/catalyst-dev:create-pr	In Review
/catalyst-dev:merge-pr	Done

Customize state names via stateMap in your project config.

Ticket Detection

Skills detect tickets automatically from plan frontmatter (ticket: PROJ-123), filenames, handoff documents, and worktree directory names.

Linear ⇄ GitHub Sync

Catalyst’s feedback routing (see Feedback Config) prefers Linear but falls back to a GitHub issue on a configured repository when Linear is unavailable. Maintainers can mirror those GitHub issues back into Linear via Linear’s native GitHub integration, so all auto-filed tickets land in the same triage queue regardless of who filed them.

Setup (one-time, Linear workspace admin):

1. In Linear, open Settings → Integrations → GitHub.
2. Connect the Linear workspace to the repository that receives fallback filings (default: coalesce-labs/catalyst, or whatever catalyst.feedback.githubRepo is set to in your project config).
3. In the connector’s issue-sync rules, filter on the auto-submitted label so only agent-filed issues are mirrored.
4. Map the target Linear team (e.g., CTL) and the default status (e.g., Backlog).

Once configured, any GitHub issue created by a Catalyst skill surfaces in the maintainer’s Linear workspace automatically, preserving the auto-submitted label plus the skill-name label (e.g., oneshot, orchestrate). See Linear’s GitHub integration docs for the current setup UI.

Sentry

Production error monitoring via the catalyst-debugging plugin.

Setup: npm install -g @sentry/cli + add org, project, and authToken to secrets config.

Context cost: ~20K tokens when enabled. Enable only during debugging:

/plugin enable catalyst-debugging    # +20K context
/plugin disable catalyst-debugging   # -20K context

Related skills: /catalyst-debugging:debug-production-error, /catalyst-debugging:error-impact-analysis, /catalyst-debugging:trace-analysis

Research agent: @catalyst-dev:sentry-research (Haiku) — gathers error data via Sentry CLI.

PostHog

Product analytics via the catalyst-analytics plugin.

Setup: Add apiKey and projectId to secrets config.

Context cost: ~40K tokens when enabled. Enable only when analyzing user behavior:

/plugin enable catalyst-analytics    # +40K context
/plugin disable catalyst-analytics   # -40K context

Related skills: /catalyst-analytics:analyze-user-behavior, /catalyst-analytics:segment-analysis, /catalyst-analytics:product-metrics

Exa

Optional web search and code-search augmentation for research agents via the Exa MCP server.

Setup: Add exaApiKey to secrets config. Used automatically by @catalyst-dev:external-research when doing web/library research.

Context cost: MCP server — small baseline, per-query token cost.