Industrial Platform Research

Industrial Platform Agent Tools

Open-source integration toolkit for connecting AI agents, MCP clients, autonomous workflows, agent orchestrators, and developer applications to Industrial Platform research and data services.

Industrial Platform is building specialized machine-callable services that other agents can delegate work to instead of rebuilding research, evidence gathering, synthesis, validation, and related capabilities inside every workflow.

Current service

AI Web Research API for Agents

A pay-per-result research service for current, source-backed web research with citations, structured evidence gathering, uncertainty handling, and automated quality review.

Apify Actor

industrial_platform/research-brief-agent

Public Actor

https://apify.com/industrial_platform/research-brief-agent

Hosted MCP endpoint

https://mcp.apify.com?tools=industrial_platform/research-brief-agent

Use it when an agent needs:

• current web research
• competitor research
• vendor comparison
• SaaS comparison
• software pricing research
• product comparison
• market research
• technology research
• current-fact verification
• multi-source synthesis
• research with citations
• source-backed decision support

Why this repository exists

The long-term goal is not to serve one manually operated agent at a time.

Industrial Platform is being built as infrastructure that can sit underneath:

• autonomous agents
• multi-agent systems
• MCP clients
• agent frameworks
• agent orchestrators
• workflow products
• AI applications
• developer platforms
• machine-to-machine systems

A single integration should be capable of producing many downstream calls.

This repository provides the public integration layer for that model.

developer / framework / orchestrator              ↓     Industrial Platform tools              ↓       specialized service              ↓        structured result              ↓calling system continues its workflow

Quick start

The fastest current integration path is through Apify's hosted MCP infrastructure.

MCP endpoint

https://mcp.apify.com?tools=industrial_platform/research-brief-agent

This exposes the Industrial Platform research Actor as an MCP-accessible tool for compatible clients.

See:

examples/mcp/README.md

for the current MCP integration guide.

Machine contract

The current research service accepts one required field and two optional fields.

Required

research_question

The question or research task to investigate.

Example:

{  "research_question": "Compare the current pricing and major features of Notion, ClickUp, and Asana for a five-person small business."}

Optional

context

Background that helps define the user's situation, constraints, intended use, or decision.

requirements

Instructions controlling scope, comparison criteria, output format, sources, calculations, citations, or other deliverables.

Complete example

{  "research_question": "Compare HubSpot CRM, Pipedrive, and Zoho CRM for a small business.",  "context": "A five-person company needs predictable pricing, automation, integrations, and straightforward administration.",  "requirements": "Use current authoritative sources. Include a comparison table, distinguish monthly and annual pricing, cite sources, identify important limitations, and state unresolved facts."}

Output contract

A completed research result contains:

research_questionresearch_datebriefstatusqaresearchusage

A customer-ready result has:

status = ready

brief

The completed source-backed research deliverable.

qa

Automated quality-review metadata.

Typical contents include:

scoreverdictissuessummaryrevision status

research

Research metadata such as:

evidence countunresolved questions

usage

Model-request and token-usage metadata.

Research workflow

The current service follows a multi-stage pipeline.

research question      ↓current web research      ↓structured evidence packet      ↓uncertainty handling      ↓research writing      ↓automated QA      ↓revision when required      ↓customer-ready result

Evidence research

The research stage searches current web sources and attempts to prioritize authoritative and primary evidence where available.

Uncertainty handling

Missing, conflicting, inaccessible, or insufficiently verified information should remain explicitly unresolved rather than being silently invented.

Research writing

The customer-facing brief is produced from the collected evidence packet and supplied user context.

Automated QA

Completed work is reviewed for:

• requirements coverage
• factual and analytical correctness
• evidence discipline
• formatting and instruction compliance
• clarity and usability

Delivery gate

The current research Actor uses a quality threshold before a result is treated as customer-ready.

QA score >= 85ANDverdict = PASS

Good delegation targets

A calling agent should consider delegating to Industrial Platform when the task requires substantially more than a trivial lookup.

Examples:

Need current competitor pricing        ↓delegate research

Need evidence from several sources        ↓delegate research

Need a vendor comparison with citations        ↓delegate research

Need unresolved facts identified explicitly        ↓delegate research

A simple lookup may be cheaper and faster for the calling agent to perform directly.

Tool-selection heuristic

A parent agent can think of delegation approximately as:

internal search cost+ search latency+ synthesis cost+ verification cost+ expected error risk
>
cost of delegating the task

When the delegated service is cheaper or operationally simpler, the parent agent can call Industrial Platform and continue its workflow with the returned result.

MCP integration

MCP is currently the primary open developer-integration path in this repository.

Hosted endpoint

https://mcp.apify.com?tools=industrial_platform/research-brief-agent

Actor

industrial_platform/research-brief-agent

Intended MCP flow

AI application      ↓MCP client      ↓Industrial Platform research tool      ↓Apify execution      ↓source-backed research result      ↓calling application continues

Detailed setup:

examples/mcp/README.md

REST and API integration

The research Actor can also be invoked programmatically through Apify's Actor API.

Planned examples in this repository include:

examples/curl/examples/python/examples/javascript/

These examples will demonstrate:

• authenticated execution
• input submission
• result retrieval
• status handling
• error handling
• machine-readable output consumption

Do not commit API tokens, wallet credentials, secrets, or private keys to this repository.

Authentication

Paid Actor execution requires an authenticated execution path.

Depending on the integration, authentication may be handled through:

• Apify account authentication
• supported API-token flows
• supported MCP authentication
• supported agentic-payment infrastructure

Credentials should be stored in environment variables, secret managers, or the calling platform's secure credential store.

Never hardcode secrets into:

README filesexample scriptssource filesconfiguration committed to Git

Payments and machine transactions

Industrial Platform is being designed for both conventional developer usage and machine-to-machine transactions.

The current research product uses pay-per-result execution through Apify.

The completed research billing event is:

research-brief-completed

A result that does not pass the delivery threshold should not be treated as a completed research result for that event.

Calling systems should inspect the live Actor metadata for current pricing and execution eligibility before invoking paid services.

Agentic-payment availability depends on the underlying platform, supported payment infrastructure, and current service eligibility.

Developer integration strategy

Industrial Platform is being designed so developers can integrate once and create many downstream service calls.

Target integration surfaces include:

Apify StoreApify APIApify MCPMCP clientsagent frameworksworkflow enginesJavaScriptPythonRESTagentic-payment systemsA2A-compatible discoverymachine-readable manifestsdeveloper registries

The objective is to make Industrial Platform services available wherever agent developers choose tools.

Repository structure

Current and planned structure:

industrial-platform-agent-tools/│├── README.md├── LICENSE├── .gitignore│├── examples/│   ├── mcp/│   │   └── README.md│   ├── curl/│   ├── python/│   └── javascript/│├── packages/│   └── research-mcp/│├── manifests/│   ├── server.json│   └── agent-card.json│└── docs/    ├── quickstart.md    ├── machine-contract.md    ├── pricing.md    ├── mcp.md    └── agentic-payments.md

Not every planned path is implemented yet.

The repository will expand as integrations are completed and validated.

Planned distribution layers

1. Apify

Current execution and monetization infrastructure.

2. MCP

Direct tool exposure to compatible AI systems and developer applications.

3. JavaScript

Simple integration examples and package support for JavaScript and TypeScript environments.

4. Python

Simple integration examples for Python-based AI and automation systems.

5. Official and downstream registries

Machine-readable metadata intended to make Industrial Platform services discoverable through developer and agent-tool registries.

6. Agent-to-agent discovery

Planned machine-readable agent metadata for systems that delegate tasks between remote agents or services.

Planned service tiers

The current production service is the deeper research workflow.

Future services may include a lower-latency machine-oriented research tier designed for high-frequency agent workloads.

Potential distinction:

FAST RESEARCH- narrower evidence search- shorter output- fewer research passes- lower latency- lower transaction price- high-frequency machine usage

DEEP RESEARCH- broader evidence gathering- more extensive synthesis- richer comparison output- automated QA- revision when needed- complex research tasks

Any future tier will be benchmarked for:

• latency
• evidence quality
• QA pass rate
• model cost
• platform cost
• failure rate
• gross margin
• developer usefulness

before being treated as a production offering.

Scale objective

The intended architecture is designed for repeatable machine demand rather than one-off manual usage.

one developer integration        ↓many downstream agents        ↓repeated specialized requests        ↓Industrial Platform services        ↓paid machine transactions

The scaling problem is therefore primarily a distribution and integration problem.

Industrial Platform aims to reduce developer friction until adding a specialized external capability is easier than recreating that capability inside every agent.

Example developer use cases

Agent framework

An agent framework can expose Industrial Platform research as an optional external research capability.

user task   ↓framework planner   ↓research required?   ↓Industrial Platform   ↓cited result   ↓framework continues

Procurement workflow

vendor shortlist   ↓research competing vendors   ↓compare current pricing and features   ↓return cited evidence   ↓procurement logic

Market-intelligence workflow

company / market question   ↓current web research   ↓source reconciliation   ↓structured brief   ↓analysis pipeline

Autonomous application

agent detects evidence gap   ↓calls Industrial Platform   ↓receives verified research result   ↓continues original task

Reliability principles

Industrial Platform integrations should prefer explicit uncertainty over unsupported certainty.

Services should aim to:

• prefer authoritative sources where available
• preserve source URLs
• distinguish facts from analysis
• identify unresolved evidence gaps
• avoid fabricating missing information
• preserve meaningful caveats
• expose useful execution metadata
• fail visibly rather than silently when appropriate

Automated QA reduces error risk but does not guarantee factual perfection.

Security

Do not commit:

• API keys
• access tokens
• private keys
• wallet seed phrases
• passwords
• account credentials
• customer secrets

Use environment variables or secure secret-management systems.

If a credential is accidentally committed, revoke or rotate it immediately.

Contributing

This repository is currently maintained by Industrial Platform.

Issues and pull requests may be used for:

• integration fixes
• compatibility improvements
• documentation corrections
• example implementations
• developer experience improvements
• protocol support
• framework integrations

Service-specific production infrastructure may remain separately maintained.

License

Integration examples and open-source code in this repository are provided under the MIT License unless otherwise noted.

See:

LICENSE

Industrial Platform

Industrial Platform builds agent-native research, analysis, and data-processing services for:

• autonomous AI agents
• developers
• agent frameworks
• MCP clients
• multi-agent systems
• workflow products
• machine-to-machine applications

The goal is to make specialized external capabilities easy for other systems to discover, evaluate, invoke, pay for, and consume programmatically.

Organization

industrial-platform-ai

Current research Actor

industrial_platform/research-brief-agent

Developer toolkit

industrial-platform-agent-tools