invokit-mcp
OfficialClick on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@invokit-mcp search for tools to generate images"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
invokit-mcp
Official MCP server for invok.it — the API marketplace for AI agents.
Search, evaluate, and invoke AI tools directly from Claude Desktop, Cursor, VS Code, or any MCP-compatible client.
Quick Start
Add the following to your MCP client configuration:
{
"mcpServers": {
"invokit": {
"url": "https://api.invok.it/mcp",
"headers": {
"Authorization": "Bearer ik-your-api-key"
}
}
}
}Get your API key at invok.it/dashboard/api-keys.
Claude Desktop
Open Settings → Developer → Edit Config
Paste the config above (replace
ik-your-api-keywith your key)Restart Claude Desktop
Cursor
Open Settings → MCP
Click Add new MCP server
Paste the config above
VS Code
Add to .vscode/mcp.json in your project:
{
"mcpServers": {
"invokit": {
"url": "https://api.invok.it/mcp",
"headers": {
"Authorization": "Bearer ik-your-api-key"
}
}
}
}Related MCP server: disvr
Available Tools (14)
Discovery
Tool | Description |
| Semantic search for AI agent tools |
| Get full details for a tool by slug |
| List all categories with counts |
| Search multi-tool workflow skills |
| Get skill details with tool bindings |
| Search deterministic step-pipeline apps |
| Get app details with execution steps |
Invocation
Tool | Description |
| Invoke a tool via transparent proxy (requires API key) |
Quality & Feedback
Tool | Description |
| Quality score breakdown (success rate, schema honesty, latency) |
| Performance metrics (latency percentiles, error rate, uptime) |
| Find type-compatible alternative tools |
| Submit quality feedback after invoking a tool (improves rankings) |
Platform
Tool | Description |
| Check your current API quota, usage, and remaining calls |
| Platform capabilities, endpoints, and rate limits |
Recommended Workflow
search_tools → get_tool_quality → invoke_tool → submit_outcomeSearch for tools matching your task
Evaluate quality scores to pick the best tool
Check your usage quota with
check_usageInvoke the tool
Submit quality feedback via
submit_outcometo improve rankings
Available Tools
14 toolscheck_usageA
Check your current API usage, quota, and remaining calls for this month.
Requires an API key (INVOKIT_API_KEY environment variable). Use this before invoking tools to ensure you have remaining quota.
Inputs: no named parameters. Returns: result
get_appA
Get full details for a specific app by its slug.
Args: slug: The app's unique identifier.
Inputs: slug (required). Returns: result
get_manifestA
Get the invok.it platform manifest — capabilities, endpoints, auth methods, and rate limits.
Inputs: no named parameters. Returns: result
get_skillB
Get full details for a specific skill by its slug.
Args: slug: The skill's unique identifier.
Inputs: slug (required). Returns: result
get_toolB
Get full details for a specific tool by its slug.
Args: slug: The tool's unique identifier (e.g. "send-email-via-sendgrid").
Inputs: slug (required). Returns: result
get_tool_alternativesA
Find type-compatible alternative tools from the same category.
Useful when a tool's circuit breaker is open or you want to compare options.
Args: slug: The tool's slug identifier. limit: Maximum number of alternatives to return (default 5).
Inputs: slug (required), limit. Returns: result
get_tool_metricsA
Get raw performance metrics for a tool — invocation counts, latency percentiles (p50/p95/p99), error rate, and uptime. Use this for detailed performance analysis.
For a simple reliability check with an overall score, use get_tool_quality instead.
Args: slug: The tool's slug identifier. period: Time period — "hour", "day", "week", "month", or "all_time".
Inputs: slug (required), period. Returns: result
get_tool_qualityA
Get the quality score breakdown for a tool — an overall 0-100 score with primary signals (success rate, schema honesty, latency consistency) and secondary signals. Use this to evaluate whether a tool is reliable enough to use.
For detailed latency percentiles and raw performance data, use get_tool_metrics instead.
Args: slug: The tool's slug identifier. period: Time period — "hour", "day", "week", "month", or "all_time".
Inputs: slug (required), period. Returns: result
invoke_toolA
Invoke a tool on the invok.it marketplace via transparent proxy.
Requires an API key (INVOKIT_API_KEY environment variable).
Check the tool's input_schema first (via get_tool) to know what arguments to pass. The response includes the tool's output along with platform metadata (latency, tool version, invocation ID if available).
After receiving the result, use submit_outcome to provide quality feedback on the tool. This helps improve tool rankings for all agents.
On failure, error responses include an error code (e.g. CIRCUIT_OPEN, RATE_LIMIT_EXCEEDED, TOOL_NOT_FOUND), a suggestion for resolution, and alternative tools when available (on 503). Use this information to automatically recover — switch to an alternative tool or retry after the suggested delay.
Args: slug: The tool's slug identifier (e.g. "send-email-via-sendgrid"). arguments: JSON arguments matching the tool's input_schema (use get_tool to see the schema). version: Specific tool version to invoke (e.g. "1.0.0"). Uses latest if omitted.
Inputs: slug (required), version, arguments. Returns: result
list_categoriesA
List all tool categories with counts of tools, skills, and apps.
Inputs: no named parameters. Returns: result
search_appsB
Search for apps (deterministic step pipelines) on invok.it.
Args: query: Natural language search query (e.g. "document processing", "image analysis pipeline", "ETL workflow"). category: Filter by category slug. pricing_type: Filter by "free" or "paid". tags: Comma-separated tags to filter by. page: Page number. per_page: Results per page.
Inputs: page, tags, query, category, +2 more. Returns: result
search_skillsA
Search for skills (multi-tool workflows) on invok.it.
Args: query: Natural language search query (e.g. "pdf to summary", "data pipeline", "content moderation"). category: Filter by category slug. pricing_type: Filter by "free" or "paid". tags: Comma-separated tags to filter by. page: Page number. per_page: Results per page.
Inputs: page, tags, query, category, +2 more. Returns: result
search_toolsA
Search the invok.it marketplace for AI agent tools using natural language.
Args: query: Natural language search query (e.g. "send email", "image recognition"). category: Filter by category slug (use list_categories to see available ones). min_quality_score: Minimum quality score (0-100). max_price_per_call: Maximum price per API call in USD. page: Page number for pagination. per_page: Results per page (max 100).
Inputs: query (required), page, category, per_page, +2 more. Returns: result
submit_outcomeA
Submit quality feedback after invoking a tool.
This closes the feedback loop: search → invoke → report outcome. Outcomes improve quality scores and help other agents choose better tools.
Requires an API key (INVOKIT_API_KEY environment variable).
Args: slug: The tool's slug identifier that was invoked. invocation_id: The invocation ID from invoke_tool's response metadata. agent_id: A stable identifier for your agent (default: "mcp-client"). Use a custom value like "my-chatbot-v2" to track feedback by agent. semantic_accuracy: How accurately did the output match the expected result? 0.0 = completely wrong, 0.5 = partially correct, 1.0 = perfect. output_usefulness: How useful was the output for your task? 0.0 = useless, 0.5 = somewhat helpful, 1.0 = exactly what was needed. task_completion: Did the tool complete the intended task? feedback_text: Optional free-text feedback (max 2000 chars).
Inputs: slug (required), invocation_id (required), agent_id, feedback_text, +3 more. Returns: result
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 14 tools
Every tool targets a distinct action and resource: quality scores, raw metrics, alternatives, tool details, skills, apps, categories, invocation, and feedback are all clearly separated. The two most similar tools, get_tool_quality and get_tool_metrics, explicitly cross-reference each other to eliminate ambiguity.
Tool names consistently follow a verb_noun snake_case pattern across get, search, list, check, submit, and invoke. Entity types like tool, skill, app, and category appear in predictable positions, making the naming scheme easy to learn and anticipate.
14 tools is well within the ideal range for a marketplace platform server. Each tool covers a necessary step in the workflow: discover, inspect, check usage, invoke, give feedback, and retrieve platform metadata. No tool feels redundant or extraneous.
The tool surface covers the full marketplace lifecycle: searching and retrieving tools/skills/apps/categories, inspecting quality and raw metrics, finding alternatives, checking quota, invoking tools, and submitting outcome feedback. The only possible gap would be direct invocation of skills or apps, but those are not advertised as invocable resources.
Maintenance
Related MCP Connectors
MCP server for progressive tool usage at any scale (see https://klavis.ai)
Find MCP servers, APIs, tools and skills by intent. Free key: POST /api/agents/register
Pay-per-use tool marketplace for AI agents. Search, price-check, and call APIs via MCP.
Official MCP server for Agentwork — delegate tasks to AI agents with human-in-the-loop
Related MCP Servers
- AlicenseBqualityCmaintenanceMCP (Model Context Protocol) Server for skill4agent - Search, view, and install AI skills in AI conversations.332 npm4MIT
- AlicenseNot gradedqualityDmaintenanceTool search engine for AI agents. One API call to discover the best MCP server for any task. 900+ services indexed with 4-dimensional value ranking.MIT
- AlicenseAqualityDmaintenanceMCP server that exposes 300+ AI agents as tools via a single API key. Supports listing agents, invoking any agent with chat-completion style messages, checking agent health, and retrieving platform statistics.54MIT
- AlicenseAqualityDmaintenanceAn MCP server that gives AI coding agents real-time access to the public-apis dataset, enabling search, live probing, and code snippet generation for 1,500+ free public APIs.75MIT