MCP Server

The MCP server lets AI clients query and manage your rewards data in natural language. It exposes the same read/write operations as the REST API as callable tools. Endpoint: https://mcp.rewards.so

Connecting Claude

Claude.ai + Claude Code — simplest

Go to claude.ai → Customize → Connectors and click Add connector. Fill in:

Field	Value
Name	`Rewards`
URL	`https://mcp.rewards.so`

Save. An OAuth popup will ask you to authorize your Rewards account. Once approved, the connector is available in all Claude.ai conversations. Claude Code bonus: if you’re signed into Claude Code with the same account, connectors you’ve added on claude.ai automatically appear in Claude Code — no extra setup needed.

Claude Code (CLI, alternative)

If you prefer to add it directly from the terminal:

claude mcp add --transport http https://mcp.rewards.so

The OAuth flow opens in your browser. If the auto-redirect fails, copy the callback URL from your browser’s address bar and paste it back into the terminal.

Claude Desktop

Claude Desktop uses a separate config file — it does not sync with your claude.ai account. Edit ~/Library/Application Support/Claude/claude_desktop_config.json on macOS, or %APPDATA%\Claude\claude_desktop_config.json on Windows:

{
  "mcpServers": {
    "rewards": {
      "url": "https://mcp.rewards.so"
    }
  }
}

Restart Claude Desktop. A browser window will open to authorize access on first launch.

All clients support OAuth natively via RFC 7591 Dynamic Client Registration — no pre-configured credentials needed.

Connecting other clients (Cursor, etc.)

Clients that don’t support OAuth can authenticate with an API key. Generate one in Account → Integration, then pass it as a Bearer token:

{
  "rewards": {
    "url": "https://mcp.rewards.so",
    "headers": {
      "Authorization": "Bearer rw_yourkey"
    }
  }
}

Keys with read permission enable read-only tools; keys with all permission enable read and write tools. If your key has access to multiple point systems, add:

X-Point-System-Id: <your-point-system-uuid>

Available tools

Every tool takes an optional pointSystemId, required only if your key covers more than one point system.

Read tools

Tool	Description
`list_point_systems`	The point systems this key can reach
`get_point_system`	Metadata and stats (total wallets, total points, ban count)
`get_leaderboard`	Top wallets ranked by points. Params: `limit` (1–200), `page`
`get_wallet`	Points, rank, events, and blacklist status for an address or ENS name
`search_wallets`	Search wallets by address substring. Params: `query`, `limit`
`list_wallets`	Page through every wallet with its points and blacklist status. Params: `page`, `limit`, `search`, `sort`
`get_wallet_events`	Full point history for a wallet
`get_blacklisted_wallets`	List of banned wallets with reasons
`get_analytics`	Wallets, points and events over a window, each compared with the window before it. Param: `days`
`get_activity_log`	Who did what, 30-day retention. Params: `page`, `limit`, `actionType`
`list_point_triggers`	Point triggers: a trigger key your app fires, worth a fixed number of points
`list_reward_automations`	Reward automations: conditions that queue wallets into a rolling draft reward
`get_automation_options`	The views, levels and event types you can build automation conditions from
`list_rewards`	Reward campaigns with their status and claim counts
`get_reward`	One campaign in detail, including claimed vs unclaimed
`list_views`	Saved leaderboard filters, referenced by `view_rank` conditions
`list_badges`	Badge families and their tiers (progression levels), with each tier’s conditions. The “Levels” family is the migrated legacy levels
`search_docs`	Search the Rewards documentation (docs.rewards.so). Params: `query`, `limit`
`get_doc`	Read one documentation page in full. Param: `slug`

Write tools

Write tools require an all API key or OAuth with write scope.

Tool	Description
`add_points`	Add points to a wallet (creates it if new). Params: `address`, `amount`, `eventType` (optional)
`add_points_bulk`	Same thing for many wallets in one call. Param: `entries`
`ban_wallet`	Blacklist a wallet. Params: `address`, `reason`
`unban_wallet`	Remove a wallet from the blacklist. Params: `address`
`create_point_trigger`	Create a point trigger. Params: `name`, `triggerKey`, `pointsAmount`
`update_point_trigger`	Rename a trigger or change what it is worth
`toggle_point_trigger`	Activate or pause a point trigger
`delete_point_trigger`	Delete a point trigger
`create_reward_automation`	Create a reward automation from conditions and a reward template
`toggle_reward_automation`	Activate or pause a reward automation
`delete_reward_automation`	Delete a reward automation
`create_reward_draft`	Draft a reward campaign
`update_reward_draft`	Edit a campaign that is still a draft
`delete_reward_draft`	Delete a campaign that is still a draft
`create_view` / `update_view` / `delete_view`	Manage saved leaderboard filters
`create_badge_family` / `update_badge_family` / `delete_badge_family`	Manage badge families
`create_badge_tier` / `update_badge_tier` / `delete_badge_tier`	Manage tiers (progression levels) and their conditions

All tools accept ENS names (.eth) in addition to 0x addresses.

What the MCP will never do

Publishing a campaign on-chain, cancelling one, and upgrading the contract all need a wallet signature, so they only exist in the dashboard. The MCP stops at the draft: an agent can prepare a campaign down to the last wallet, but you are the one who signs it. update_reward_draft and delete_reward_draft refuse anything that is no longer a draft or already has an on-chain id.

Example prompts

"Analyse le top 50 du leaderboard et identifie les wallets qui semblent farmer"
"Ban tous les wallets avec moins de 3 events distincts et plus de 10 000 points"
"Montre-moi l'historique complet de 0x1234..."
"Ajoute 500 points à vitalik.eth pour l'event discord_launch"
"Quels wallets ont été bannis cette semaine ?"
"Compare les 30 derniers jours aux 30 précédents"
"Crée un trigger connect_telegram à 50 points"
"Prépare un draft de campagne pour le top 100, 0.001 SOMI par wallet"
"Qui a modifié les automations ce mois-ci ?"

Audit trail

Every write is recorded in the activity log with the API key or OAuth client ID. All MCP actions are fully auditable from the dashboard, or through get_activity_log.

Getting started

Dashboard

Connecting Claude

Claude.ai + Claude Code — simplest

Claude Code (CLI, alternative)

Claude Desktop

Connecting other clients (Cursor, etc.)

Available tools

Read tools

Write tools

What the MCP will never do

Example prompts

Audit trail

​Connecting Claude

​Claude.ai + Claude Code — simplest

​Claude Code (CLI, alternative)

​Claude Desktop

​Connecting other clients (Cursor, etc.)

​Available tools

​Read tools

​Write tools

​What the MCP will never do

​Example prompts

​Audit trail

Connecting Claude

Claude.ai + Claude Code — simplest

Claude Code (CLI, alternative)

Claude Desktop

Connecting other clients (Cursor, etc.)

Available tools

Read tools

Write tools

What the MCP will never do

Example prompts

Audit trail