MCP server for IG Harness — operate an Instagram account from AI agents over MCP
This server has 29 tools with significant definition gaps. Most tools lack output schema documentation, critical for agent planning. Descriptions are present but generic (avg ~60 chars, well below the 194-char baseline). Parameter descriptions exist but lack constraints (enums, ranges, formats). No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite 20 WRITE tools. Error handling is absent, no recovery guidance or actionable error messages. Tool naming is verb-first (good), but composition issues exist: send_dm + broadcast could be unified; manage_* tools are overloaded (create/update/delete in one tool). No evidence of idempotency guarantees on destructive operations. Security: no visible rate limiting, permission gates, or audit logging in tool definitions.
Get a summary of account statistics and active scenarios
Send a broadcast message to multiple followers
Apply engagement gates to multiple reels at once
Create a rule for automatically responding to comments
Create a form for collecting follower information
Create a media post (photo/video) for publishing
Create a new automation scenario for DM workflows
Create a tracked link for measuring engagement
No output schemas documented for any tool. Agents cannot plan downstream calls or extract required fields (e.g., does list_followers return follower_id for use in send_dm?). This violates pattern:tool and pattern:tool-chain.
10 manage_* tools combine create/update/delete in one tool via action enum. This violates pattern:tool (one responsibility per tool) and forces agents to reason about which action to invoke. Split into separate tools: create_*, update_*, delete_*.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 49 | 2026-07-28+ | v2 |
Enroll a follower in an automation scenario
Get detailed information about a specific follower
Retrieve submissions from a form
Get click analytics for a tracked link
List CRM objects (followers, scenarios, tags, etc.)
List followers of the Instagram account
List recent Instagram Reels from the account
Create, update, schedule, or delete broadcast campaigns
Update or delete comment auto-response rules
Create, update, or delete engagement gates for gating content
Manage follower tags, notes, and metadata
Update or delete forms
Manage LINE account connections and integrations
Update, publish, or delete media posts
Create and manage rich message templates
Update, activate, deactivate, or delete automation scenarios
Create, update, or delete staff members with role-based access
Create, update, or delete tags for organizing followers
Update or delete tracked links
Send a direct message to an Instagram user
Upload an image for use in messages or forms
20 WRITE tools (send_dm, broadcast, create_*, manage_*) lack tool annotations (destructiveHint, idempotentHint). Agents cannot determine if a call is safe to retry or has irreversible consequences. Add destructiveHint=true for delete/send operations.
Parameter descriptions lack constraints. E.g., 'trigger_type' in create_scenario says 'Type of trigger (e.g., message, follow)' but does not declare valid enum values. 'limit' in list_followers has no range (1 - 100? 1 - 1000?). Add enums and min/max to descriptions.
No error handling guidance. Tools lack recovery hints (e.g., 'User not found. Try list_followers() first.'). No categorization of errors as retryable vs. fatal. Agents cannot self-correct on failure.