projio

Project knowledge orchestrator and MCP server for research repositories.

Projio turns a research repository into a queryable knowledge environment for humans and AI agents. It layers structured, machine-accessible knowledge over a repo — code, papers, notes, pipelines, figures, questions — and exposes it through a unified MCP server so humans and AI agents work from the same view.

Install

pip install projio                # core orchestrator + MCP server
pip install "projio[all]"         # all ecosystem packages

What projio owns vs. what it coordinates

Projio itself is small: workspace scaffolding, docs-site workflows, the MCP entrypoint, and the questio research-question layer. Everything else lives in focused sibling packages that projio composes. This separation keeps each domain independently usable and specced.

The engineering vs. science split is deliberate — see the delegation model for how pipeio flows, result notes, questio questions, and deliverables each own distinct content and link in one direction.

Ecosystem

Retrieval substrate

Package	Goal	Spec
indexio	Domain-agnostic corpus indexing, chunking, embedding, and semantic search — the shared retrieval infrastructure every other package registers sources with	site

Knowledge layer

Package	Goal	Spec
biblio	Project-centric bibliography: Zotero sync, OpenAlex/Crossref enrichment, PDF fetch, GROBID/Docling parsing, compiled BibTeX	bib architecture
notio	Structured project notes, idea capture, worklog — hosts the manuscript and present subpackages for paper and deck production	site
codio	Code-library registry with three tiers (core / shared / external), reuse discovery, implementation-strategy intelligence	code tiers

Engineering layer

Package	Goal	Spec
pipeio	Agent-facing pipeline authoring — flows, contracts, notebook lifecycle, Snakemake integration. Owns engineering, not science	pipeio spec
figio	Declarative figure orchestration: FigureSpec YAML → panel rendering → SVG composition → PDF/PNG	site

Science & delivery

Subsystem	Goal	Spec
questio (in projio core)	Research questions, hypothesis tracking, prior art, binding of questions to results and deliverables	questio
notio/manuscript	Section assembly, citation checking, figure insertion, pandoc → PDF/LaTeX	site
notio/present	Slide decks from reusable sections — reveal.js and Marp backends, cross-project section import	presentio
deliverables	Narrative artifacts for external audiences: reports, decks, posters — bind questions and results into a story	deliverables

Which subpackage do I need?

I want to...	Use
Find existing code, notes, or papers by semantic search	indexio
Manage a project's bibliography, fetch PDFs, resolve citekeys	biblio
Capture an idea or record a project decision	notio (notes)
Discover reusable code or register a new library	codio
Author a data-processing pipeline or notebook	pipeio
Build a figure from a declarative spec	figio
Track a research question and bind results to it	questio
Assemble a manuscript for submission	notio/manuscript
Build a slide deck or reveal.js talk	notio/present

Key capabilities

• Search before creation — discover existing implementations, consult literature, then decide: reuse, wrap, or implement new
• 70+ MCP tools — unified agent interface across all subsystems, scoped to the current project
• Three workspace kinds — generic, tool, and study scaffolds for different project types
• Engineering/science separation — pipeio builds the machinery; questio, result notes, and deliverables carry the findings and narrative
• Documentation site — MkDocs Material with monorepo plugin and semantic-search chatbot integration

Documentation

The docs follow the Diataxis structure, plus a long-form Handbook for the agentic research workflow:

Section	Purpose	Start here
Handbook	Narrative textbook for the open-science research stack (BIDS, DataLad, Snakemake, Marimo, Quarto/MkDocs, projio, agentic on top); companion to the September 2026 workshop	Why this stack, Outline
Tutorials	End-to-end guided paths	Quickstart
How-to guides	Task-focused recipes	Initialize a workspace
Explanation	Design choices and concepts	Ecosystem, Delegation model
Reference	Command and layout details	CLI, MCP tools