grainulator
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
What it adds
Good models still need a way to keep track of evidence, revisit assumptions, and verify their work. Grainulator gives them that process, with a local ledger that stays with the task as it moves between models and sessions.
Capability | What it gives you |
Evidence that stays with the work | Typed claims, sources, evidence tiers, and provenance that survive edits and exports. |
A check on the answer | Conflict detection, weak-support signals, and explicit gaps to investigate. |
Clear next actions | Two lists: what the agent can continue automatically, and what needs your input. |
Sessions you can carry forward | Configurable research, stop/resume, reusable context, and credential-free exports. |
Verification for your task | A managed command loop that can use a verifier you provide. |
Grainulator works with Claude Code and Codex through plugins. It also provides a command-line tool and an MCP connection for other agents. You do not need to install its components separately.
Related MCP server: Ouroboros
Quick start
Install the CLI and matching native runtime through the shared Homebrew tap:
brew tap grainulation/homebrew https://github.com/grainulation/homebrew
brew install grainulation/homebrew/grainulator
grainulator doctor
grainulator demoHomebrew provides Node 24 and the matching native assets. The offline demo needs no credentials. For agent integration, follow agent setup.
From source
Requires Node.js 24 or later; Node 24 and 26 LTS are the tested release lines. Node 24 is the development default.
Upgrading from an earlier version? Follow the migration checklist to update the plugin and keep your existing research data.
git clone --branch v3.0.1 https://github.com/grainulation/grainulator.git
cd grainulator
npm ci --ignore-scripts
node bin/grainulator.js doctorTo use Grainulator inside your agent, follow the Claude Code or Codex setup guide.
Start a portable research session without making provider calls:
grainulator research --question "What evidence answers this question?" --dir ./research-session --prepare-onlySee CLI research for provider configuration and checkpoint continuation.
Grainulator 3: one CLI/plugin with a unified configuration and separately verified native runtime assets. Follow installation, upgrading, configuration and data retention. The shared Homebrew tap lists available stable formulas.
Other agents: connect through MCP
Add an MCP server that runs this command, with your project's absolute path:
node /absolute/path/to/grainulator/bin/grainulator.js mcp --dir /absolute/path/to/projectAntigravity installs the full plugin. See the agent setup guide for Antigravity, Cursor and other MCP agents.
For the full Claude Code or Codex plugin, use the agent setup guide. Codex needs a project folder; the guide shows how to set it. CLI and direct MCP access remain available to other hosts.
Or point your coding agent (Claude Code, Codex, or Antigravity) at this checkout:
Read README.md and AGENTS.md in the Grainulator repository. Run node bin/grainulator.js doctor
and node bin/grainulator.js demo, and report the exit codes and final status lines. Then tell me
which routes below this machine can run, and ask me before starting any model-backed or paid run.Route | What it does | Needs | Time / Cost |
1. Local ledger & offline demo | Runs | Node.js ≥24; no credentials | ~2 s · $0 (no model calls) |
2. Managed command verifier | Runs | Route 1 + local verifier script | Per-round timeout & round cap |
3. Host plugin or MCP server | Connects | Route 1 + host agent | Host session · $0 Grainulator overhead |
4. Semantic & adversarial evals | Runs | Route 1 (+ API credentials for live model matrix) | Offline $0 · live matrix requires explicit budget |
Put it to work
For necessary-premise links, current source checks, recorded verification runs,
audit findings and missing acceptance criteria, see
evidence maintenance. Start with
node bin/grainulator.js maintain --help; all operations use an explicit sprint.
With Grainulator connected, give your agent the outcome you need:
Use Grainulator to investigate whether we should migrate this service. Record the evidence, challenge the assumptions, implement the agreed changes, and verify them. Keep the remaining next steps split into Auto and Manual.
The work follows a simple cycle:
Record findings with their sources and evidence tiers.
Check for conflicts, weak support, and missing perspectives.
Act on the gaps that matter, then verify the requested result.
Continue from the saved ledger or export the session to another workflow.
There are no fixed claim-count or research-pass quotas. Auto lists work the agent can continue within your authorization. Manual lists decisions, access, or actions that need you. When you ask only for next steps, those two lists are the entire response.
node bin/grainulator.js init --dir ./sprints/migration \
--question 'Should we migrate this service?' \
--audience engineers --constraints 'Preserve existing user data' \
--done 'A verified plan with remaining risks documented'
node bin/grainulator.js add --dir ./sprints/migration \
--id r001 --type constraint --topic migration \
--content 'Existing user data must remain readable.' --evidence stated
node bin/grainulator.js compile --dir ./sprints/migrationThe equivalent MCP tool is add_claim on the grainulator server. See the tool reference for evidence, memory, export, analytics, and orchestration operations.
Explore the docs
Guide | Start here for… |
Tagged source, isolated archives, and verified build identity. | |
Host updates, renamed tools, existing data, and retired setup instructions. | |
Use Grainulator with Claude Code, Codex, or another agent. | |
Models, feature controls, credentials, export, and resume. | |
The canonical CLI and MCP interface. | |
Attaching a model command and a task-specific verifier. | |
The workspace layout and internal modules. | |
What has been measured and what remains unproven. | |
Development setup, tests, and changes to the project. |
Verification and limits
CI checks Node 24 and 26, isolated package installation, plugin contracts, Rust runtime conformance, lint, and credential-free CLI/MCP research and export workflows. Native host and live-provider acceptance have separate verification guides.
The compiler checks the structure and support recorded in the ledger; it does not establish that a source is true. Evaluations do not establish a general accuracy or efficiency gain, and per-pass limits are not a whole-session spending cap. Keep those distinctions when interpreting results.
npm test
npm run test:install
npm run lintBrowser and runtime prerequisites are in CONTRIBUTING.md. The offline node bin/grainulator.js demo exercises the adapter/verifier protocol without a provider account.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
Evidence-backed project memory and checkpoints for AI coding agents.
Source-backed prompting and coding-agent guidance with provenance and freshness checks.
One memory for AI agents in Claude Code, Cursor, Codex and ChatGPT that flags conflicting facts.
AI-native programming language and toolchain for accountable, verifiable agent-driven development.
Related MCP Servers
- FlicenseNot gradedqualityCmaintenanceA shared, concurrency-safe ledger for LLM code audits. It enables multiple agents to collaboratively report, verify, and resolve bugs with deduplication and hallucination resistance.-
- AlicenseBqualityBmaintenanceLocal-first Agent OS that wraps Claude Code, Codex CLI, and other coding agents in a replayable Seed → Ledger → Runtime contract, driven by an interview → seed → execute → evaluate → evolve workflow loop.3429,739 PyPI6,385MIT
- AlicenseNot gradedqualityAmaintenanceEnables evidence-gated, multi-session AI coding runs with plan-build-ship state management, coordinating Claude Code and Codex native agents.20 npmMIT
- AlicenseAqualityBmaintenanceEnables coding agents to analyze a repository evidence-first, answering questions as verifiable claims with source evidence, targeted test observations, and reference resolution.53Apache 2.0