Grainulator plugin
Grainulator keeps an evidence ledger for research and engineering work and checks a model answer for conflicts and gaps.
- 86 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The plugin runs local commands and a managed verification loop with your verifier.
- The research playground spends limits of the connected model provider.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/grainulatorDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The plugin runs local commands and a managed verification loop with your verifier.
- The research playground spends limits of the connected model provider.
This is third-party code. Review the repository files before installing.
What it does
Grainulator adds a working ledger to Claude Code and Codex: claims with sources, evidence tiers and provenance. The ledger stays local and follows the task between models and sessions. The plugin looks for conflicts between claims and weakly supported spots, then ends with two lists of next steps. Auto holds what the agent continues under your authorization, Manual holds decisions and access that need a person. The same toolset is available as a command-line tool and as an MCP server for other agents.
Who it is for. Researchers and engineers running a long task with an agent who want to see what each conclusion rests on.
Good fit when
- Before a service migration you collect facts, challenge assumptions and only then change code.
- A study runs for several days and the conclusion must continue in a new session without retelling context.
- A model answer reads smoothly but it is unclear which claims have sources behind them.
- The agent result is checked by your own verifier script in a managed loop.
Not a fit when
- The task is short and one-off, so an evidence ledger only slows it down.
- The team cannot install Node.js 24 or newer.
- You want a ready package from the npm registry: release 2.2.0 ships through GitHub.
Example request
Use Grainulator to investigate whether we should migrate this service. Record the evidence, challenge the assumptions and split next steps into Auto and Manual.Limitations
Node.js 24 or newer is required. Version 2.2.0 ships as a GitHub release, not through the npm registry, so installation runs from tagged source. The research playground started by `preview` runs models through a provider account you configure locally and spends its limits. Codex needs a project folder for the plugin, and the agent setup guide shows how to set it. Moving from 1.x needs a separate plugin migration, and existing research data is kept.
How to disable. Remove the plugin in Claude Code or Codex and drop the MCP entry if you added one. Clean the research ledger folders by hand.
FAQ
How do I install Grainulator in Claude Code or Codex?
Clone tag `v2.2.0`, run `npm ci --ignore-scripts` and `node bin/grainulator.js doctor`, then follow `docs/AGENT-SETUP.md` for your agent.
How do I connect it to other agents such as Cursor?
Add an MCP server with the command `npx -y github:grainulation/grainulator#v2.2.0 mcp --dir` and the absolute project path. Cursor and other MCP agents connect the same way.
What are Auto and Manual in the answer?
Auto lists work the agent continues under your authorization. Manual lists decisions, access and actions that need you.
What should I do when upgrading from 1.x?
Follow the migration checklist in `docs/UPGRADING.md`: it covers the plugin update and keeping existing research data.
How do I record a claim from the terminal?
Initialize a session with `node bin/grainulator.js init`, add a claim with `add` giving type, topic and evidence tier, then run `compile`.
Discussion
No questions or reviews yet.
Related
PluginsA Claude Code plugin that makes the model finish the job: result verification, evidence and systematic investigation enforced as procedure
A Claude Code plugin that runs research on language model internals: papers, hypotheses, experiments, verification, a report
A Claude Code plugin with hooks and 26 skills: measures what the agent knows, blocks edits until code is read, keeps memory
Plugin and MCP server that asks GPT, Gemini, Grok and local models to review a plan or hunt for bugs before code ships.
A plugin that syncs Claude Code, Codex CLI and Antigravity on one project and lets them hand work to each other.
A plugin and MCP server that let Claude Code drive the local Guaardvark AI studio on your own graphics card
README in short
The plugin, command-line tool and MCP server live in one project. The website shows a saved evidence-check example and does not send your question to a model.