Give agents and AI products precise context from code, specs, Office files, and PDFs. Ragmir indexes locally, on a laptop, a VPS, or an offline server, and returns exact citations. Your agent answers with the model you choose.
Ask Claude Code or Codex to build a feature from project documents. The agent searches, reads, refines, then codes.
001·Use cases
Four workflows. One cited evidence layer.
Build features and fix incidents with a coding agent, ground an AI product in customer documents, or decide in a confidential local chat.
Build a feature with Claude Code or Codex.
Your coding agent searches the specification, checks exceptions, then changes code and tests.
01Trigger
Ask Claude Code or Codex to add account recovery.
02Sources
The selected Word specification and authentication ADR.
03Retrieval
The agent finds rules, opens sources, and searches for exceptions.
04Cited result
The agent updates code and tests, citing the rules it applied.
Diagnose an incident before changing production code.
Your coding agent grounds a regression fix in the runbook, incident report, and relevant source.
01Trigger
Ask Claude Code or Codex to diagnose a checkout timeout.
02Sources
The timeout runbook, incident report, and retry implementation.
03Retrieval
The agent verifies retry rules and opens the cited implementation.
04Cited result
The agent fixes the regression and adds a test from the evidence.
Ground an AI product in your customer's documents.
A TypeScript app or a LangChain.js chain retrieves cited passages on the customer's own server, and refuses when the documents are silent.
01Trigger
A user asks the contract assistant for the termination notice period.
02Sources
The customer's contracts and policies, indexed on their own server.
03Retrieval
RagmirRetriever returns the exact clause with its page reference.
04Cited result
The app answers with the cited clause, or says the documents do not cover it.
Make an architecture decision with a confidential chat.
Discuss internal constraints with a chat and model running locally or on infrastructure you operate.
01Sources
Internal architecture notes and constraints on attachment storage.
02Retrieval
Ragmir finds the storage requirements and their sources.
03Local model
The selected model runs on this machine or on your own server.
04Cited result
A cited decision: files in object storage, metadata in PostgreSQL.
Your own documents, too.
The same cited retrieval works on a personal corpus: the research behind a video, a lifetime of paperwork, or your whole career history. Indexing and search stay on your machine.
01 / 03 Content creation
Write YouTube scripts from your research
Index papers, notes, and your past transcripts. Your agent drafts the script and cites a source under every claim, so fact-checking takes minutes.
[1] research/hybrid-search-paper.pdf:p4#2
02 / 03 Personal admin
Handle paperwork with a local model
Index bank statements, payslips, contracts, and certificates on your own machine. A local model assembles each application and points to the exact page as proof.
[1] payslips/2026-09.pdf:p1#0
03 / 03 Career
Search for a job with your career history
Index every CV, annual review, and project report. Match each offer against your own evidence and write cover letters that invent nothing.
[1] reviews/2024-annual-review.pdf:p2#1
002·Install
Install once. Retrieve through CLI, TypeScript, or MCP.
Add Ragmir Core to the repository that owns the files, using Node.js 22.12 or later. Choose the sources, then ingest. Local-hash retrieval needs no account, API key, or model download.
Set up Ragmir
Let your coding agent tailor the setup, or choose your package manager and run the commands yourself.
Let your coding agent configure Ragmir
Paste this prompt into your coding agent. It inspects the repository, asks before making choices, installs the right packages, and verifies cited retrieval.
003·Integrations
Plugins for Claude Code, Codex, Cursor, and Gemini CLI. A retriever for LangChain.
Use Ragmir from the tools your team already runs. Every integration reads the same local index and returns the same exact citations: install it as a coding-agent plugin, a Gemini CLI or Qwen Code extension, Agent Skills, a LangChain.js or Python retriever, a Docker image, or an MCP server for any client.
Your agent chooses when to search, expand a citation, or refine a query. MCP starts with at most three compact citations by default, names the knowledge base it serves, and returns passages already seen as short stubs. The agent handles reasoning, generation, and actions.
Connect Claude Code, Codex, Kimi, OpenCode, or Cline
The same cited retrieval, whichever application consumes it.
01 / 03
On your machine
Run Ragmir, your chat or agent, and a downloaded model on the same machine, for example with Ollama. Keep inference local and disable remote calls in the consumer when passages must stay on that machine.
02 / 03
On your own server
Use a model you host on a server or private cloud infrastructure. Retrieved passages travel to that server. You manage who can access them, the connection, and retained logs.
03 / 03
With a cloud provider
Connect a compatible chat or coding agent to Ragmir through MCP or your application. Passages sent to the model leave your environment and are subject to its provider’s data policy.
005·For teams shipping RAG
From a customer folder to cited answers, on their infrastructure.
Forward deployed engineers and AI-native teams stand up retrieval where the documents live: a laptop for the first demo, then a VPS, an on-premise server, or a network without internet access.
01 / 04
A data path you can sign
Parsing, embeddings, the index, and search run in one Node.js process. No embedding API, telemetry, or hosted store to declare in the security review.
02 / 04
Proof on their documents
rgr evaluate scores retrieval on the customer’s own questions, checks exact citations, and counts hard negatives that must return nothing.
03 / 04
Runs on their servers
Debian or Alpine containers, a non-root user, a read-only root filesystem, no network. The index lives on a volume and a scheduled job keeps it fresh.
04 / 04
A clean handoff
One committed ragmir.config.json, rgr doctor for diagnosis, and upgrades that keep the last validated index until its replacement passes.
Your service keepsTLS · authentication · per-user access · rate limits · the model
One library for project knowledge, with document parsing and OCR included. Add semantic embeddings when your corpus needs them.
01 / 03 Included
Project retrieval
Turn selected files into cited context for agents, scripts, and TypeScript applications.
Prepare
Run
Index changes incrementally, refresh after Git operations, and flag citations whose file changed.
Reach the code that defines what you ask about, plus specs and Office rows with cited headers.
Expand a cited passage, verify its indexed version, and spot weak evidence.
Use the same local index through CLI, MCP, or TypeScript.
02 / 03 Optional
Scans, photos, and legacy files
Keep the evidence in scanned PDFs, photos, and legacy Word files searchable with local extractors.
Prepare
Run
Extract embedded text first; OCR only blank pages.
Keep page citations attached to retrieved passages.
Reuse cached OCR results in bounded batches.
Detect Tesseract, OCRmyPDF, textutil, antiword, or LibreOffice.
03 / 03 Optional
Semantic retrieval
Add meaning-based matching with a local embedding model when keyword evidence is not enough.
Prepare
Run
Install Transformers only when you need it.
rgr models list recommends a model for this machine; the download stays explicit.
Combine lexical and vector evidence with source filters.
Measure retrieval quality with your own golden queries.
Supported formats
More than 900 file types are indexed locally, each with a citation agents can quote. Design files, mind maps, mail, APIs, data models, and diagrams are cited object by object.
OpenAPI AsyncAPI Postman Insomnia Bruno HAR GraphQL Protobuf Avro Thrift Smithy TypeSpec WSDL .http
Data and configuration
CSV JSON JSONL YAML TOML XML INI Jupyter GeoJSON SARIF SPDX Logs
Translations
gettext .po XLIFF Apple .strings String Catalog .resx ARB Fluent
BI and reporting
Tableau Power BI (TMDL) LookML SSRS SSIS JasperReports SAS Stata Nextflow Snakemake
Code and infrastructure
TypeScript JavaScript Python Java Kotlin Swift Go Rust C# C/C++ PHP Perl Ruby Dart Scala Elixir Haskell Julia Zig Solidity COBOL ABAP Terraform Bicep Helm Nix Dockerfile Shell .NET projects WiX Godot Luau Lean Gherkin Robot Framework XSLT Apple plist
With an optional local extractor
PDF OCR PNG JPEG TIFF WebP HEIC .doc
007·Control your data
Local retrieval. An explicit data path.
Ragmir keeps its index on the machine where it runs and returns source text without masking. Confidentiality also depends on the chat or agent, the model host, and their logs and network settings.
01 / 04
Documents on disk
Select the folders Ragmir can read. Ragmir does not upload your source files to a hosted document store.
02 / 04
Git-ignored state
The index, reports, and generated integrations stay in the local git-ignored .ragmir/ folder.
03 / 04
Selected sources
Include useful folders and exclude credentials, generated output, and unrelated data.
04 / 04
Explicit consumption
Your application decides where retrieved passages go. A cloud model provider receives those you include in its requests; self-hosting requires access, transport, and logging controls.
008·FAQ
Frequently asked questions
Sources, local models, OCR, and integration with your development tools.
What is Ragmir?
Ragmir is the retrieval layer for agentic RAG: it indexes selected project files and returns exact cited passages through TypeScript, CLI, or MCP. Your agent decides when to search again, reasons over the evidence, and generates the answer. Default retrieval works offline without a model download.
Does Ragmir send my confidential code or documents?
Ragmir has no telemetry or hosted document store. It indexes on the machine where it runs and returns unmasked passages. Your chat or agent can forward those passages to a model provider. For confidential work, control the whole path: client, local or self-hosted model, connection, access, and logs. Optional semantic setup explicitly downloads model weights.
Which AI agents and tools does Ragmir work with?
Ragmir Core is model-agnostic. Claude Code, OpenAI Codex, and Cursor install it as a plugin, and Gemini CLI and Qwen Code as an extension, each with the retrieval skill and the MCP server. Kimi, OpenCode, and Cline connect through rgr setup. LangChain.js applications use RagmirRetriever or the agent tools, and Python applications install ragmir, langchain-ragmir for LangChain and LangGraph, or llama-index-retrievers-ragmir from PyPI. Any agent that loads skills installs the Ragmir skills with npx skills add jcode-works/ragmir, and coding assistants read the documentation on Context7. Other MCP clients, scripts, and CI jobs get the same cited results. Each server reads one project's base and warns or refuses unrelated workspaces.
Can I use Ragmir on personal documents, such as bank statements or my CV?
Yes. Index the folders you choose: research and past transcripts for a YouTube script, a lifetime of paperwork, or your whole career history for a job search. Indexing and search run on your machine. Pair Ragmir with a local model such as Ollama to keep every passage there; with Claude Code, Codex, or Claude Desktop, the cited excerpts go to that provider.
How does a team share the same knowledge base?
Commit a shared ragmir.config.json and share reviewed source documents through your normal Git or file-sharing workflow, then run rgr ingest on each workstation, or let rgr hooks install refresh the index after each pull. Each developer keeps a local index. Ragmir does not manage branches or synchronize peers.
Does Ragmir require an API key, a model download, or an internet connection?
The default local-hash provider works offline with no API key or model download. Semantic retrieval needs the optional Transformers package and a model preload. A fully local workflow also uses a local consuming model; Ollama is one option. To reuse a model you already serve, the opt-in openai-compatible provider embeds through Ollama, vLLM, or another OpenAI-compatible endpoint.
Can I see where ingestion time and memory are spent?
Optional ingestion metrics report phase timings, throughput, queues, OCR cache hits, and memory use. These local diagnostics omit document content and queries; there is no usage-tracking service.
How does Ragmir differ from IDE code indexing or hosted RAG?
Native code search handles structured code well. Ragmir targets project documents that need citations, keeps its index local, and exposes bounded passages through non-destructive MCP tools. Small output budgets keep exact citations and return typed summaries with omission counts.
Does Ragmir write the answers for me?
Ragmir supplies the search and citation tools for agentic RAG. Your agent, chat, or application chooses when to call them and sends the evidence to its model. The integration guide covers local Ollama, self-hosted servers, and cloud providers.
Which file formats can Ragmir index?
More than 900 file types index with no extra install: source code in over 250 languages, text, Markdown, PDF, Word, PowerPoint, and Excel (macro-enabled files and templates included), OpenDocument, EPUB, HTML and saved web pages, CSV, JSON, YAML, translations, and build files such as Dockerfile. Exports are cited object by object: Figma, Sketch, Excalidraw, and tldraw frames; XMind sheets; emails, mailboxes, and Evernote notes; Postman and Insomnia requests and HAR captures; OpenAPI and AsyncAPI operations; PowerDesigner, ERwin, MySQL Workbench, draw.io, Visio, BPMN, ArchiMate, and UML models; Tableau and Power BI projects; GeoJSON features and SARIF findings. Scanned PDFs, photos, and legacy Word files can use optional local extractors, and unsupported files are flagged explicitly.
Can Ragmir run on a VPS or a server without internet access?
Yes. Ragmir runs wherever Node.js 22.12 runs, including Debian and Alpine containers as a non-root user with a read-only filesystem and no network. It opens no HTTP port: agents reach it through stdio MCP, and your service embeds the TypeScript client and owns authentication, authorization, rate limits, and TLS. The official Docker image runs the CLI and MCP server without installing Node.js, and the server guide includes a working Dockerfile.
Can I ship Ragmir inside a customer deployment or a commercial product?
Yes, under the AGPL-3.0 terms, which include offering the source of a modified version you distribute or run for users over a network. To integrate Ragmir into proprietary software or a hosted service without those obligations, JCode Works offers a separate commercial license.
Start with the coreAGPL-3.0 open source
Give your agents the context your project needs.
Give your agents and AI products cited evidence through MCP, CLI, or TypeScript, on a laptop or on your servers. Keep the model and workflow you choose.