Skip to content

README.md

Latest commit

 

History

History
223 lines (167 loc) · 12.9 KB

README.md

File metadata and controls

223 lines (167 loc) · 12.9 KB
Raw

Confidence

Confidence AI Plugin

Official Confidence plugin for AI coding tools. Manage feature flags, experiments, onboarding, and migrations right from your agent.

Docs License Version

✨ Highlights

  • Manage feature flags without leaving your editor: create, list, update, target, resolve, and archive flags from Claude Code, Cursor, Codex, or Gemini CLI
  • One-command migrations from PostHog, Eppo, Statsig, or Optimizely (flags and SDK code; Optimizely also migrates users, teams, roles, policies, and Flag clients)
  • Guided onboarding: spin up a Confidence account, invite teammates, create SDK clients, and connect a warehouse without reading a setup doc
  • Docs on tap: search Confidence documentation and SDK integration guides inline while you code
  • Works with every major AI coding assistant through the same MCP servers

ℹ️ Overview

Confidence is Spotify's feature flagging and experimentation platform, built on the OpenFeature standard. This plugin exposes Confidence's flag management, documentation, and migration tooling as MCP servers and slash commands, so you can create a flag, plan a migration off another platform, or onboard a new workspace directly from your AI coding assistant's chat.

⬇️ Installation

Claude Code

claude plugin install confidence

Cursor

From the Marketplace: open Cursor Settings → Plugins, search for Confidence, and click Install.

Manual setup: add the MCP servers to .cursor/mcp.json in your project (or ~/.cursor/mcp.json globally):

{
  "mcpServers": {
    "confidence-flags": {
      "url": "https://mcp.confidence.dev/mcp/flags"
    },
    "confidence-docs": {
      "url": "https://mcp.confidence.dev/mcp/docs"
    }
  }
}

Codex

codex plugin marketplace add spotify/confidence-ai-plugins
codex
/plugins
# Select Confidence and install

Gemini CLI

gemini extensions install https://github.com/spotify/confidence-ai-plugins

Local Development

git clone https://github.com/spotify/confidence-ai-plugins.git
claude --plugin-dir ./confidence-ai-plugins

Cursor: File → Open Folder on this repository (the confidence-ai-plugins directory itself — not Desktop or your home folder). Skills load from ./skills/ via .cursor-plugin/plugin.json and from .cursor/skills/ (symlinks) for project-level discovery. Commands load from ./commands/ (including /migrate-optimizely plan access, plan flags, execute code, etc.) and from .cursor/commands/ (symlink to ../commands). Dedicated /migrate-optimizely-* menu shortcuts also exist in Cursor. A parent folder as the workspace will not auto-load them.

Skills only, any agent

The migration and onboarding skills can also be installed individually via the skills CLI, which supports Claude Code, Cursor, Codex, Gemini CLI, and 70+ other agents:

npx skills add spotify/confidence-ai-plugins

This installs only the skill logic (auto-triggering guidance for migrations and onboarding). It does not configure the confidence-flags/confidence-docs MCP servers or the /confidence:* slash commands, even when targeting one of the four clients above. Since several skills call into those MCP servers to manage flags, use one of the full installs above for complete functionality; use this only if you want the skill guidance on an agent the full installs don't support.

🚀 Usage

Once installed, just ask your assistant:

> List my feature flags
> Create a flag called new-checkout with a boolean schema
> /confidence:onboard-confidence create-account
> /confidence:migrate-posthog plan flag
> /confidence:migrate-posthog plan code
> /confidence:migrate-eppo plan flag
> /confidence:migrate-eppo plan code
> /confidence:migrate-statsig plan flag
> /confidence:migrate-statsig plan code
> /confidence:migrate-optimizely plan access
> /confidence:migrate-optimizely adjust access
> /confidence:migrate-optimizely execute access
> /confidence:migrate-optimizely plan flags
> /confidence:migrate-optimizely adjust flags
> /confidence:migrate-optimizely execute flags
> /confidence:migrate-optimizely plan code
> /confidence:migrate-optimizely adjust code
> /confidence:migrate-optimizely execute code
> /confidence:analyze-project

Features

This plugin provides access to Confidence tools across these categories:

  • Feature flags: Create, list, update, archive, resolve, and target feature flags
  • Onboarding: Create accounts, invite users, set up SDK clients, configure warehouses, and learn experimentation concepts
  • Documentation: Search Confidence docs and SDK integration guides
  • Migration: Migrate feature flags from PostHog, Eppo, Statsig, or Optimizely to Confidence. Optimizely Phase 0 also maps access (users, groups, roles, policies, Flag clients) — see Optimizely access
  • Project analysis: Scan a project and propose meaningful feature flag opportunities tailored to the codebase

Slash Commands

  • /confidence:onboard-confidence <create-account | invite-user | create-client | setup-wizard | setup-warehouse | learn | status>: Create accounts, onboard users, set up SDK clients, configure warehouses, and learn experimentation concepts
  • /confidence:migrate-posthog <plan flag | plan code | execute <plan-file>>: Migrate feature flags from PostHog to Confidence
  • /confidence:migrate-eppo <plan flag | plan code | execute <plan-file>>: Migrate feature flags from Eppo to Confidence
  • /confidence:migrate-statsig <plan flag | plan code | execute <plan-file>>: Migrate feature flags from Statsig to Confidence
  • /confidence:migrate-optimizely <plan access | adjust access | execute access | plan flags | adjust flags | execute flags | plan code | adjust code | execute code | execute <plan-file>>: Migrate Optimizely Feature Experimentation to Confidence — see Optimizely access (Flag clients are handled in plan access Step 4; Cursor also exposes dedicated / menu shortcuts)
  • /confidence:analyze-project [project-dir]: Analyze a project and propose meaningful feature flag changes using Confidence

Optimizely → Confidence

PostHog, Eppo, and Statsig migrate flags then code. Optimizely also migrates access (Phase 0) before those. Agent contract: skills/migrate-optimizely/SKILL.md (including Adjust Access / Flags / Code: Steps) and IAM mapping in skills/migrate-optimizely/access.md.

Phase 0 — Access (users, groups, roles, policies, clients)

Writes a plan first. Nothing is invited or created in Confidence until you tick consent and run execute.

Command What it does Writes to Confidence?
plan access / -plan-access (also bare /migrate-optimizely) Map Optimizely collaborators and teams to Confidence invites, groups, intended shares, and Flag-client candidates (Step 4). File: .claude/plans/optimizely-access-migration-<date>.md. When the plan is complete, the agent must ask what next (adjust / tick consent / execute / done) — there is no automatic path into adjust. If SDK keys arrive later, re-run plan access No
adjust access / -adjust-access Fine-edit that plan in chat: users, groups, roles, policies, clients (e.g. “skip all @example.com”, “Checkout should be Editor”). You do not have to hand-edit the markdown. Also offered from the plan-access exit ask No
execute access / -execute-access After you tick [x] Invite / [x] Create: create groups + Reader policies, send invites, create ticked Flag clients, then provision each person as they accept (group, policy, client, flag shares) Yes (IAM)

What adjust access can change

Kind Examples
Users Invite or skip (one email, a team, a domain, or all); move between groups; add an email you provide
Groups Create or skip; rename; merge/split; change members
Roles Viewer / Editor / Owner shares per group or user; override default mapping (e.g. Publisher → Viewer)
Policies Group policy roles (default Reader); yes/no on tightening default-policy
Clients Create or skip; names; split/merge; which groups see which clients

What it will not do: invent people or Flag clients; flatten teams into per-user shares; map Project Owner to workspace Admin; put Flags Editor/Reader on a workspace policy; change admin-policy except to add known Administrators; invite or create resources during plan/adjust; delete a live group/user because a row is now Skip.

Silence is not consent. Tick every user and group (and listed Flag clients) before execute. Invites last 7 days. Flag clients need environment SDK keys — if the export has none, that step is skipped until keys exist.

Phase 1 — Flags

Same plan → exit ask → optional adjust → execute pattern as access. Running experiments and partial-% rollouts stay out by default (different bucketing).

Command What it does Writes to Confidence?
plan flags / -plan-flags Scan Optimizely flags and write .claude/plans/optimizely-flag-migration-<date>.md. Lists flags with no Optimizely rules as auto everyone catch-all. Must run a Rules operator audit: on exists / substring, tell the user and auto-map exists → IS NOT NULL, substring → starts with / ends with. regex and mid-string contains stay BLOCKED. When complete, the agent must ask what next (adjust / tick / execute / done) No
adjust flags / -adjust-flags Fine-edit that plan: scope, Migrate/Skip ticks, client, bucketing, schema, rules. Also offered from the plan-flags exit ask No
execute flags / -execute-flags (1) create flag shells → (2) suggest rules import → (3) suggest resolve-verify ALL flags (segment match) — this validates Phase 1 → then plan code. Auto everyone catch-all if a flag still has zero enabled rules. Never call Phase 1 done after rules alone Yes (flags)

Phase 2 — Code

Command What it does Writes to Confidence / repo?
plan code / -plan-code Scan Optimizely SDK usage and write .claude/plans/optimizely-code-migration-<date>.md. When complete, the agent must ask what next (adjust / execute / done) No
adjust code / -adjust-code Fine-edit that plan: style, resolve mode, transforms, files/flags. Also offered from the plan-code exit ask No
execute code / -execute-code Transform code / open PRs from the plan (typically one PR per flag) Yes (repo)

MCP Servers

Server Endpoint Description
confidence-flags https://mcp.confidence.dev/mcp/flags Feature flag management
confidence-docs https://mcp.confidence.dev/mcp/docs Confidence documentation

Supported Clients

Client Config Marketplace
Claude Code .claude-plugin/ Official plugin
Cursor .cursor-plugin/ Cursor Marketplace
Codex .codex-plugin/ Via marketplace command
Gemini CLI gemini-extension.json Direct from repo

Documentation

💭 Community & Support

Found a bug or have a feature request? Open an issue. Changes are tracked in the changelog.

License

Apache License 2.0