MCP server for Reddit Ads API v3 - enables access to Reddit advertising accounts, campaigns, ad groups, ads, and performance reporting
Reddit Ads MCP exhibits solid fundamentals with clear verb-driven naming, reasonable descriptions, and visible parameter schemas. However, significant gaps emerge in output schema documentation, error handling guidance, and parameter constraints. All 6 tools have descriptions (10 - 19 chars above the minimum), and most parameters include type annotations via C# attributes. The main weaknesses are: (1) no documented output schemas for any tool, LLMs cannot plan downstream tool chains; (2) lack of enums/constraints on open-ended string parameters (accountId, campaignId, fields, breakdowns); (3) no error handling guidance or recovery paths; (4) date format validation is implicit, not documented; (5) no pagination support despite list tools potentially returning large datasets. This is competent but not production-grade.
Get daily performance for the last N days (default 7). Convenience wrapper that returns impressions, clicks, spend, CTR, CPC, eCPM broken down by DATE and CAMPAIGN_ID.
Get a performance report for a Reddit ad account. Returns fields like impressions, clicks, spend, CTR, CPC, and eCPM broken down by the specified breakdowns.
List all Reddit ad accounts accessible with the current credentials. Discovers accounts via /me/businesses and then fetches ad accounts for each business.
List ad groups for a Reddit ad account, optionally filtered by campaign. If accountId is not provided, uses the default account.
List ads for a Reddit ad account, optionally filtered by ad group. If accountId is not provided, uses the default account.
List all campaigns for a Reddit ad account. If accountId is not provided, uses the default account from REDDIT_ACCOUNT_ID.
No output schemas documented for any tool. LLMs cannot see what fields to expect (e.g., does ListCampaigns return campaign_id, name, status, budget?) and cannot chain tools. Without documented outputs, agents cannot build multi-step workflows.
Parameter 'fields' and 'breakdowns' in GetPerformanceReport accept arrays with free-form string values. No enum constraint or validation. LLM might pass 'IMPRESSON' (typo), 'REVENUE' (unsupported), or arbitrary strings. Should declare: fields: enum[IMPRESSIONS, CLICKS, SPEND, CTR, CPC, ECPM], breakdowns: enum[DATE, CAMPAIGN_ID, AD_GROUP_ID, ...].
accountId parameter appears in 5 of 6 tools as optional with default from REDDIT_ACCOUNT_ID env var. While this reduces friction, the fallback is silent, no explicit mention in descriptions that an unset REDDIT_ACCOUNT_ID will cause failure. If the env var is missing, error message will be opaque. Should validate at startup and document the dependency.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 46 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 32 | - | v1 |
Date parameters (startDate, endDate in GetPerformanceReport) accept strings with implicit format 'YYYY-MM-DD' and silently normalize to ISO 8601. Code shows NormalizeDate() appends 'T00:00:00Z', but this is not documented in parameter descriptions. LLM might pass '2024-01-15T12:30:00Z' expecting specific time precision, not realizing it gets forced to midnight UTC.
List tools (ListCampaigns, ListAdGroups, ListAds) return raw API responses via JsonHelper.Format(doc). No pagination, no result count limit documented. If a Reddit account has 500 campaigns, all 500 will be returned as JSON, bloating context window and degrading LLM reasoning. Should add limit/offset parameters and cap at 50 items per call.
No error handling guidance. If Reddit API returns 401 Unauthorized (token expired), 404 (campaign not found), or 429 (rate limited), the tool will fail with the raw HTTP error. Descriptions do not explain what to do on failure, should document: 'Returns an error if accountId does not exist. Verify the account is accessible via ListAccounts.'
campaignId and adGroupId parameters are optional and unvalidated. If an LLM passes a non-existent ID, the filter silently returns an empty result or an error from the API. Tool descriptions do not advise LLM to call ListCampaigns first to discover valid IDs.