Research Toolkit MCP Server

Research Toolkit

English · 简体中文

Develop long-form research in English or Chinese from a question or existing material, with concrete judgments and traceable sources.

Use it for long-form research and writing across fields, including industry, companies, technology, policy and science. Start a new study, revise an existing report against its sources, or update conclusions when new material arrives. Agents operate Toolkit to research, write and independently review; the report serves its specified readers and purpose. Toolkit connects methods, evidence and deliverables through one research entry and optional task tools.

Choose an entry · Start research · See a research case · Documentation

How does a study get done?

The lead agent maintains the research question, effective requirements and overall judgment. Research agents investigate, compare, calculate and write substantial parts. A designated assembling agent reads their full work and sources, reconciles the argument and develops the main report progressively. Different independent agents review it, audit the findings and accept the final version.

Responsibility	How the work connects
Task and context	Keep the current brief, outline, effective requirements and dependencies together. Satisfied requirements remain visible when they still constrain later work.
Evidence and analysis	Give each researcher a bounded question, relevant original material and related findings. Preserve its evidence, reasoning, conditions and unresolved questions in the handoff.
Writing and revision	Develop meaningful research units, review premises before dependent work, and integrate their complete analysis into one owned main draft. Changes return to affected units and the outline.
Review and delivery	Complete seven separate content-review perspectives, relevant domain review, independent audit and final acceptance against the current report and sources. Resolve necessary corrections before delivery.

These are four connected responsibilities, not four fixed agents or four independent pipelines. Codex, Claude Code and similar agent tools run the agents and carry messages; Toolkit provides shared inputs, methods, artifact relationships and completion checks. Native collaboration does not require separate API setup. For script-managed execution, the shared Agent execution adapter connects to Claude Code, OpenCode or Codex, according to the current agent tool. Research steps · Handoff and calls

flowchart LR
  L[Lead: question and outline] --> R[Research agents: evidence and arguments]
  R --> W[Designated assembling agent: progressive main draft]
  R --> U[Independent review at important dependencies]
  U --> R
  W --> V[Seven content perspectives + domain review]
  V --> A[Independent audit]
  A --> F[Independent final acceptance]
  A -->|Necessary correction| R
  F --> D[Complete report]

Which should I use?

The Skill provides research methods, CLI and MCP expose the same local tools, and the plugin bundles the Skill with MCP. Choose an entry by your agent environment and need for saved task state; all use the same research and writing methods.

Your situation	What to use or install	How to start
You want your existing agent to follow the research methods	Skill: read the methods directly, or install them for repeated use. The methods themselves need neither Python nor MCP; default first-use setup is explained below.	Send the request below; the agent reads methods as needed and uses its existing research and writing capabilities.
You want methods and task tools together, and your agent tool supports plugins	Plugin: installs the Skill and MCP together; local tools need Python 3.10+ and MCP dependencies.	Follow the plugin instructions, reopen the session, select Toolkit and describe the research.
You need saved task state, recovery and links between sources and drafts, and your agent tool supports local MCP	MCP: configure the local service and use the same Skill for methods. No separate MCP setup is needed if the plugin already supplies it.	Follow the MCP setup; the agent calls research_* tools.
You work in a terminal or script, or your agent tool has no MCP support	CLI: download the repository and use Python 3.10+; no MCP dependencies needed.	The agent reads the same Skill and invokes the same task tools through the command line.

For a first use, start directly with the Skill. Add task tools when persistent state, tool-managed reviews or delivery checks help the work, not because a report reaches a word count. You do not need to install all four or use them in sequence. Skill-only use supports in-depth long reports without requiring the tools' CSV/JSONL registries or delivery receipts.

Which tasks fit? Long-form research in English or Chinese across fields, requiring multiple sources, comparison, explanation or supported recommendations. This includes developing summaries and local arguments within a long report, revising existing work and continuing it with new evidence. Simple factual questions, routine translation or minor wording edits usually need only your existing agent. The agent uses its available search and source-access capabilities to obtain and read the material.

Research from scratch · Revise or update a report · Install for your agent tool

What happens on first use? You can read the methods directly without installing a Skill or MCP. In Codex, Claude Code and OpenCode, the agent automatically reuses or downloads a complete local copy and uses the bundled Python 3.10+ script as needed to configure basic usage recording for the current project. It reuses unchanged configuration, including active plugin hooks and an existing off setting. Only added or changed configuration needs loading through the agent tool; an already-open Codex session needs a reload or reopening for those changes. The agent uses the tool's available status to check activation rather than inferring it from a saved configuration file. If Python, command access, setup or loading is unavailable or fails, it states the actual status and continues research. Basic operation events are sent by default, excluding conversations, reports and local paths; recording can be turned off. The agent explains this once on first use and handles setup rather than assigning the user a separate installation step. Setup, collected data and controls.

<a id="usage"></a>

Quick start

Describe your question or requested revision in your own words to an agent that can read the repository and sources. The example below is a starting point; you do not need to fill in undecided details. The agent uses existing context to propose an approach and clarifies only consequential gaps. You can start with the repository link; the agent handles the first-use configuration described above.

Use Research Toolkit: https://github.com/rrrrrredy/research-toolkit
Read SKILL.md and follow the methods needed for this task.

I want to understand or resolve: [The question, or what to change in an existing report.]
Report target readers and intended use: [Who will read the report and why, if known.]
Material and explicit requirements: [Sources, scope, cutoff, format or length, if available.]

Deliver a complete report that answers these questions with usable citations;
include a short change note when revising.

Start research: one usage guide

One method set supports long-form and in-depth research, revisions and updates. The request determines scope, length and depth; the agents load methods for their current work and complete independent review, audit and acceptance.

Only need to review an existing report? Use the standalone Research Review Skill. Toolkit already includes these review capabilities; existing users do not need a second installation.

Have a draft or new material? Revise or update existing research, reusing completed research and its sources.

Research case

Test-Time Compute for Reasoning Models and Agents. Compare sampling, search, revision, verification and tool feedback, then identify which operation should receive additional work. The report connects empirical scaling behavior to a method-and-evidence matrix, a worked allocation example and a design decision tree. Evidence cutoff: 9 November 2026.

Read the full report · Independent Chinese report · First draft, sources and independent reviews

The other new case examines China’s microdrama market: growth definitions, distribution models, platform competition and value-chain cash flows. It develops platform comparisons and reproducible operating scenarios, showing how traffic, rights, contracts and cash exposure shape participants’ returns.

Chinese report · English report

The two language versions independently use each question’s frozen materials; neither author receives the other-language report. The current benchmark contains 13 independent questions and 15 report assignments, retaining first outputs, located defects and revisions. Earlier office-agent, API/browser and report-update cases remain in the archive.

<a id="use-as-an-mcp-server"></a> <a id="review-and-acceptance"></a> <a id="inside-the-toolkit"></a>

Agent setup

How you want to use it	Entry
Read the methods directly	Send the request above; for files or text-only chats, see agent setup.
Install a Skill or plugin	Choose your agent · Plugin installation
Save progress and call workflow tools	MCP installation and client configuration · CLI
Try the local tools	Runnable MCP example and captured responses; synthetic examples explain the interface, while real execution records retain their actual review scope.

Nine local tools provide shared task context, on-demand methods, source and analysis saving, registered release-log queries, reviews and delivery checks. CLI and MCP use the same implementation. MCP requires Python 3.10+ and a client that can launch local processes; setup covers Windows, macOS and Linux.

Independent review uses fresh agent contexts. Each required perspective has its own reviewer; authors and assembling agents do not review their own work. Audit and final acceptance have separate assignments. CLI/MCP retains the original agent responses and input versions through the same research_review operation. Review and acceptance

<a id="evaluation-suite"></a> <a id="evaluation-results"></a>

Evidence and evaluations

Public records include complete reports, original failures, reviews and repairs. Case studies document report corrections, model comparisons cover their recorded tasks and conditions, and software checks cover tested workflow behavior. Broader quality or efficiency comparisons with native agents require comparable research tasks, inputs, models and execution conditions.

Current evidence and limits · Complete-report version comparison (Chinese) · Datasets and checks · Historical data archive

Documentation

Need	Entry
Research and writing methods	Research guide · Skill instructions
Delegate research and integrate results	Collaboration and requests · Research production and handoff
Existing reports and source updates	Continue existing research
Installation and configuration	Agent setup · Usage guide
Development and changes	Contributing · Changelog

All documentation · Website · Submit a failure case