Seneca Research Platform MCP Server

Requires a valid seneca_sk_ API key and spends credits. Commission a Seneca Research Platform study across independent AI models from leading laboratories worldwide. Use this tool to extend your analytical capabilities beyond a single model's perspective. Seneca gathers independent responses to a common research question, synthesizes their findings, and applies deterministic analysis to reveal consensus, competing positions, disagreements, minority insights, and uncertainties that may influence the conclusion. Research panels: - wide: 10 AI models across 10 laboratories in 4 countries. 15 credits. - consensus: 40 AI models across 26 laboratories in 6 countries. 45 credits. - seneca: Approximately 100 AI models across 31 laboratories in 8 countries. 100 credits. Particularly valuable for complex engineering decisions, software architecture, AI agent and MCP design, technology comparisons, strategic analysis, multi-model evaluation, LLM-as-a-Judge research, and studying how AI models interpret products, services, and descriptions for GEO/AEO optimization. When the user explicitly requests Seneca research, execute their authorized request without redundant confirmation. When independently proposing research, explain its value, recommend a panel, disclose the credit cost, and obtain authorization before spending. The server validates account eligibility, credit balance, concurrency limits, and the MCP spending ceiling of 300 reserved credits per rolling 24 hours. An unsubscribed, canceled, or past-due workspace receives subscription_required, including a 15-credit Wide run. The complimentary welcome Wide is only available in the Seneca website. This denial is a tool error; the HTTP transport may still be 200. Direct the user to https://seneca.professionalaiagents.com/pricing. insufficient_credits means an eligible subscriber needs more credits from Billing. mcp_spend_ceiling means the 24-hour MCP limit is already reserved. Provide the research prompt, selected panel, and a unique idempotency_key. The call returns a run_id immediately; research continues asynchronously. Monitor progress using get_research. Reuse the original idempotency_key, prompt, and panel when retrying an uncertain submission. Identical submissions within 24 hours return the original run without another charge. Changed inputs with the same key produce idempotency_conflict.

Requires a valid seneca_sk_ API key. Monitor the progress and completion of a Seneca multi-model research study. Seneca research panels independently collect responses from multiple AI models before synthesizing their findings into a structured, decision-oriented report. Larger panels can take several minutes to complete. Use this tool after create_research to follow the research lifecycle without interrupting execution. Returns the current research status, processing phase, participating-model progress, reserved credits, charged credits, and any reported failure_code. Research states: - queued: The research run has been accepted. - running: The research panel is being processed. - synthesizing: Seneca is preparing its findings. - completed: The research report is available. - partial: A report is available despite incomplete participation. - failed: The run ended without a deliverable report. Poll at approximately 15-second intervals, with backoff when appropriate. Continue while research is in progress. For completed or partial runs, retrieve the findings using get_research_result. Stop polling a failed run. Completed and partial reports are billable under Seneca's settlement rules. Technical failures without a report are refunded. A failure_code alone does not establish whether a refund occurred. This tool is read-only and does not consume research credits.

Requires a valid seneca_sk_ API key. Retrieve the completed Seneca Research Platform report and its structured findings. This is the primary tool for converting a multi-model research study into actionable intelligence for your user. Seneca's reports synthesize independent AI perspectives and analyze patterns of agreement, disagreement, competing positions, minority insights, and uncertainty. Depending on the study, findings may include executive conclusions, key points, supporting claims, recommendations, ranked alternatives, and evidence of differing model positions. Use this tool when get_research indicates completed or partial status. Available detail levels: - compact (default): Efficient, decision-oriented report representation, omitting the original question and placing methodology last. Recommended for most agent workflows. - full: The complete stored customer report, including additional research detail. Present the executive findings accurately, highlight meaningful disagreements, and connect conclusions to the user's original question. Preserve important uncertainty rather than treating model agreement as proof of factual correctness. Seneca measures independent model perspectives under controlled research conditions, not live web-verified facts. If research is still processing, the tool returns result_not_ready. If a run failed without producing a report, it returns research_failed. This tool is read-only and does not consume research credits.

Requires a valid seneca_sk_ API key. Examine the original research contribution from one individual AI model within a Seneca research panel. While Seneca's synthesized report provides the broader findings, this tool allows you to investigate the underlying model-level perspective in detail. Use it to inspect how a particular model explained its conclusions about the research question, review technical explanations, examine dissenting or minority positions, compare competing answers, or investigate claims that deserve additional scrutiny. Especially valuable for AI engineering research, multi-model evaluation, LLM-as-a-Judge analysis, model-behavior investigations, and understanding why the larger panel reached differing conclusions. Specify the research run_id and model_id. Responses are retrieved in bounded character windows to preserve context efficiency. The default window is 8,000 characters, with a maximum of 32,000 characters per request. If truncated is true, advance offset to retrieve the next portion. These are original research responses representing individual model perspectives. Interpret them in the context of the overall Seneca report and distinguish model assertions from independently established facts. The tool reads stored research data without making external requests. It does not consume research credits.

Requires a valid seneca_sk_ API key. Explore the individual AI models that participated in a completed Seneca research panel. Use this tool to examine the composition of the research panel, identify which models contributed successfully, and select specific model responses for closer investigation. This is especially useful when examining model diversity, laboratory representation, competing technical perspectives, minority viewpoints, or differences in model participation. Returns paginated metadata including model identifiers, display names, laboratories, response status, available token counts, and whether response text is available. Response text is intentionally excluded. Use get_model_response_text when you need to examine a particular model's original contribution. Available after research reaches completed or partial status. In-progress research returns responses_not_ready; a failed run without a report returns research_failed. Research belonging to another workspace is not accessible. This tool is read-only and does not consume research credits.

Describe Seneca Research Platform. No API key, customer data, or credits are required. Use this tool to explain what Seneca does, which research panels exist, what the panels do not do, and how a user obtains a seneca_sk_ API key. Account access and research execution use the other tools and require a valid key. This tool does not read a workspace, launch research, or reserve credits.

Requires a valid seneca_sk_ API key. Retrieve the authenticated customer's Seneca Research Platform account information, including subscription status, research eligibility, available credits, and monthly credit allocation. Use this tool to establish the customer's available research capacity before commissioning a multi-model study. It helps you recommend an appropriate research panel and determine whether sufficient credits are available. Seneca provides three research panels: Wide (15 credits), Consensus (45 credits), and Seneca (100 credits). Account eligibility and available credits determine which research options can be launched. entitled is true only when paid API or MCP research is authorized. Available credits do not override entitled false. monthly_credit_grant is 0 unless the subscription is active or trialing. A new account may hold 15 welcome credits for one Wide Research run on the Seneca website. That gift cannot be spent through MCP or the REST API. Direct an eligible free customer to https://seneca.professionalaiagents.com/. This tool is read-only, does not initiate research, and does not consume credits.