Tru8's evidence research pipeline is now available as an API and MCP server — so developers, AI agents, and automated workflows can run structured, multi-source analysis programmatically.
This post covers what the API does, how the MCP server works, and why structured evidence research is a useful capability for agent-based systems.
Why Evidence Research Needs Structure
A web search returns links. A language model returns prose. Neither gives you a structured picture of what the evidence actually looks like across multiple source types.
Tru8 sits in the gap between search and synthesis. Given a claim, it searches across 30+ source categories — government data, academic papers, news, official records, legislation, economic indicators, health data — and returns a structured evidence landscape. Every source is classified by proximity (primary, reporting, commentary) and type (data, official, news, analysis, opinion, academic). Nothing is hidden; exclusions are logged with reasons.
For developers building tools that need to reason about claims, this structure matters. It's the difference between “here are some links” and “here is the shape of the evidence, classified and mapped to the specific parts of the claim they address.”
The API
The Tru8 API provides evidence research at multiple levels of depth, so you can choose the right balance of speed and thoroughness for your use case.
Lookup
Instant retrieval of previously researched claims. If someone has already run a full analysis on a claim, you get the structured result immediately. Useful for high-volume systems that want to check before committing to a full research run.
Quick Analysis
A streamlined evidence pass that completes in roughly 15 seconds. Core retrieval and classification without the full depth of coverage recovery or extended source searching. Well-suited for real-time applications, chatbots, or agent workflows where responsiveness matters.
Full Analysis
The complete pipeline. Claims are decomposed into elements, evidence is retrieved from 30+ source categories, scored for relevance, classified by tier and type, and mapped back to the specific elements they address. The result is a full evidence landscape — the same structured output available in the Tru8 dashboard.
Smart Endpoint
A single endpoint that handles tier selection automatically. It checks for cached results first, falls back to quick analysis if speed is preferred, and escalates to full analysis when depth is needed. One call, server-side routing.
Every response includes full provenance — which sources were found, how they were classified, what was excluded (and why), and the method used for each classification decision.
MCP Server for AI Agents
Tru8 provides a Model Context Protocol (MCP) server, which means AI agents built on any supporting platform can call Tru8 as a tool — the same way they might call a calculator or a web browser.
The MCP server exposes a single tru8_check tool. An agent passes a claim and optionally a maximum tier, and receives a structured evidence landscape in return. Tier fallback is handled automatically — if a full analysis isn't available, the server returns the best available result.
This is particularly useful for agent workflows that need to:
- Ground responses in real, cited evidence rather than parametric knowledge
- Verify claims before presenting them to users
- Provide structured source breakdowns alongside generated text
- Add evidence research as a step in a multi-tool reasoning chain
- Audit the provenance of information used in automated decision-making
Because the output is structured (not prose), agents can reason over the evidence programmatically — filtering by tier, checking element-level support, or surfacing gaps where evidence is missing.
What the Response Looks Like
Every Tru8 API response returns a consistent structure, regardless of tier:
- Claims — the input decomposed into discrete, researchable elements
- Evidence — sources found, each classified by tier (primary, reporting, commentary) and type (data, official, news, analysis, opinion, academic)
- Mapping — which evidence addresses which elements, and the relationship (supports, challenges, or provides context)
- Landscape — aggregate metrics including source diversity, tier distribution, and coverage gaps
- Provenance — classification method, relevance scores, content basis, and receipts for every exclusion
This isn't a summary or a verdict. It's a structured dataset that your application or agent can interpret, filter, and present however it needs to.
Use Cases
Some of the ways developers and teams are integrating Tru8:
- AI assistants — adding evidence grounding to conversational agents, so claims are checked before being surfaced to users
- Content moderation — automated evidence checks on user-submitted claims or flagged content
- Research tools — integrating structured evidence search into academic or journalistic workflows
- Compliance and risk — verifying claims in regulatory filings, reports, or public statements
- Browser extensions — inline evidence checks on articles, social media posts, or search results
- Agent pipelines — evidence research as one step in a larger automated reasoning or decision-making workflow
Getting Started
The API is live and available now. API keys can be created from the developer portal, which also includes endpoint documentation, example requests, and response schemas.
The MCP server can be added to any MCP-compatible agent framework. Configuration details are in the developer documentation.
If you're building something that needs structured evidence research — whether it's an agent, an internal tool, or a product feature — the pipeline is there. We'd be interested to hear what you build with it.