Finsliparn
Transforms single-attempt coding into a multi-attempt, test-validated refinement loop by running your actual test suite and feeding failures back to the LLM as structured directives.
README
Finsliparn
Swedish: the honer β from finslipa, meaning "to hone", "to fine-tune" or "to put the finishing touches on."
Finsliparn is a Bun-native MCP Server and Claude Code Plugin that transforms single-attempt coding into a multi-attempt, test-validated refinement loop. It adapts the philosophy of Poetiq's ARC-AGI solver to general software development.
π― Core Value Proposition
LLMs often generate code that looks correct but fails in edge cases. Finsliparn solves this by enforcing a Refinement Loop:
- Iterate: The LLM implements a solution.
- Validate: Finsliparn runs your actual test suite (not just a syntax check).
- Feedback: Test failures are fed back to the LLM as structured, actionable directives.
- Select: If multiple attempts are made, the best one (highest score, lowest complexity) is selected.
"Test results are the objective truthβnot diffs, not prompts, not opinions."
π Architecture
Finsliparn uses a Filesystem-as-IPC architecture to support multiple AI platforms with a single core engine.
graph LR
subgraph "AI Client"
Claude[Claude Code]
Copilot[GitHub Copilot]
end
subgraph "Finsliparn Core"
MCP[MCP Server]
Directive[Directive.md]
Tests[Test Runner]
end
Claude --"Hooks"--> MCP
Copilot --"Agent"--> MCP
MCP --"Writes"--> Directive
Directive --"Reads"--> Claude
Directive --"Reads"--> Copilot
MCP --"Executes"--> Tests
directive.md: The single source of truth. It tells the LLM exactly what to do next (e.g., "Fix failing tests insrc/foo.ts").- Git Worktrees: Every iteration runs in an isolated worktree, ensuring your main branch stays clean until a solution is verified.
π Getting Started
Prerequisites
- Bun v1.3+
- Git
Installation (Development)
# Clone the repository
git clone https://github.com/jgabor/finsliparn.git
cd finsliparn
# Install dependencies
bun install
# Link the plugin (for Claude Code)
# (Coming soon)
π Documentation
- Technical Specification (v1.0.0): The core architecture and Claude Code integration.
- Copilot CLI Support (v2.0.0): Adaptation for GitHub Copilot CLI agents.
- Roadmap: Development plan and status.
π Usage
With Claude Code
# Start a refinement session
/finslipa Implement a fibonacci function in src/math.ts
Finsliparn will automatically intercept your edits, run tests, and guide you until the tests pass.
With GitHub Copilot CLI
# Start the agent
copilot run --agent finsliparn "Implement a fibonacci function"
The agent will autonomously loop through the directive.md instructions until completion.
Author
Jonathan Gabor
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.