Feixon Research Station · AI-native research workstation

An end-to-end research workstation
from reading papers to writing them.

Feixon Research Station connects literature reading, knowledge records, research agents, reusable skills and LaTeX writing into one project workspace. Its SRKB layer is a structured vector database for paper evidence, built so both AI and researchers can search, inspect and verify the same sources.

Start free ->

Paper reading · Paper notebooks · Citation-grounded AI · Collaborative LaTeX · Research skills · Team management · Community sharing

paper readingpaper notebookscitation-grounded AIsemantic searchcollaborative LaTeXresearch skillsteam managementcommunity sharingBYOK routingpaper readingpaper notebookscitation-grounded AIsemantic searchcollaborative LaTeXresearch skillsteam managementcommunity sharingBYOK routing
System

The project is built around evidence objects, not loose chat history.

Each paper becomes structured material the rest of the product can reuse: chunks for retrieval, claims for reasoning, metrics for comparison, citations for writing, and durable AI steps for long-running work.

SRKBevidence database
IngestPDFs · zip · S3
Evidence Storechunks · claims · metrics
Retrievalhybrid · pgvector
AgentsFast-RAG · research · writing
Communityshared projects · skills
Manuscriptdiffs · PDF · export
Product tour

See it work, before you sign up.

Five product surfaces that make the positioning concrete: evidence capture, semantic search, grounded chat, collaborative LaTeX writing and the open community.

Read papers with AI, keep the evidence.

Upload a paper and the AI reads alongside you — pulling out claims, results and open questions, each linked to the exact passage. You review and curate, then it is embedded into the project library, ready to query.

  • AI reading and human review on the same passages
  • Claims, results and open questions, passage-linked
  • Structural embedding into the project library
  • Runs as a background job — keep reading meanwhile
feixon.com/app/papers — reading
📄
Upload
signed upload
AI reading
claims
👤
Human review
curate
🧠
Embedding
pgvector
Ready to query
chat · search

Ask the project library, then inspect the evidence.

The research agent plans retrieval, reads stored evidence, and streams answers with citation cards and a tool-use trace you can inspect.

  • Research-agent and Fast-RAG modes
  • Hybrid or pure-vector retrieval, tunable per chat
  • Tool-use trace you can inspect step by step
  • Evidence-quality signals on retrieved material
feixon.com/app/chat

Write the paper near the evidence, together.

A multi-file LaTeX IDE with worker-side PDF builds, side-by-side preview and real-time collaboration, plus an in-editor assistant that proposes edits you can review.

  • Multi-file TeX projects with zip import/export
  • Real-time collaboration with teammates
  • PDF builds on the worker with visible job status
  • AI edit proposals as reviewable diffs
feixon.com/app/writing — main.tex
queued: latex_build_job…
✦ Assistant: tightened §4 claim & added \cite{dao2023}Review diff

Share your work, build an open research commons.

Publish your paper projects and the skills you write under an open license, so others can build on verified evidence. A provider bench ranks the most cost-effective API providers to keep token spend down.

  • Share paper projects and skills, openly licensed
  • Build on evidence others have already verified
  • Provider bench ranks the best value APIs
feixon.com/community
📦vision-lab24-paper attention pack
🧩youreview.matrix
🧩rag-evalgroundedness skill
📦bio-mlprotein-LM benchmark pack
⚖️benchprovider price ranking
Open research commonsprojects · skills · provider bench
Community

Share your work, build an open research commons.

Publish the paper projects you curate and the skills you write to the community under an open license — a small contribution to shared, rigorous research. To keep token spend down, an API provider bench ranks the most cost-effective providers for everyone.

From one paper project to a shared starting point for the field.

Publish a project and others can build on evidence you have already verified; share a skill and repeated research workflows need not be rebuilt. Everything is openly licensed.

20+shared projects
30+shared skills
10+benched providers
project.share

Share paper projects

Package a curated paper library with its extracted evidence and publish it under an open license.

skill.publish

Publish your skills

Turn a reusable research workflow into a skill others can copy into their own workspace in one click.

provider.bench

Provider value bench

An evidence-backed ranking of the API providers with the best cost, context and uptime.

copy.to.workspace

Copy to workspace

Found a project or skill you like? Copy it straight into your own project and get going.

open.license

Open licensing

Shared work carries an explicit open license — clear provenance, safe to reuse.

stars.signal

Community signal

Stars and downloads surface the projects and skills the community trusts most.

Advantage

Feixon Research Station adds the research workflow layer missing from single-purpose tools.

Search tools, source notebooks, LaTeX editors and coding agents each solve part of the job. Feixon RS’s advantage is continuity: the SRKB evidence store, AI agents, writing workspace, community and team operations share the same project context.

Research job
Common tool
Strong at
Where they stop
Where Feixon RS goes further
Find, screen and ask papers
Elicit / SciSpace / NotebookLM-style tools
Paper search, source Q&A, summaries, reports and extraction tables
Discovery and Q&A often stay separate from writing and team operations
Turns selected papers into a reusable project library for agents, search, skills and manuscripts
Write and cite papers
Overleaf / Zotero-style tools
LaTeX collaboration, references, bibliographies and source organization
Writing and references are strong, AI evidence work is usually external
Keeps LaTeX edits, PDF builds, citation-grounded AI and evidence audit together
Automate technical work
Claude Code / Codex-style agents
Codebases, scripts, files, terminals and general workflows
Research evidence objects and citation workflows must be assembled manually
Provides research-specific tools, skills and state machines around papers and manuscripts
Run a team workspace
Admin dashboards and billing tools
Roles, quotas, provider keys, job health and audit trails
Operational controls are rarely connected to paper-writing workflows
Brings roles, BYOK, quotas, job health and audit trails into the research workspace

The advantage is workflow integration: Feixon RS keeps evidence, agents, writing and operations in one project context.

Highlights

What Feixon Research Station brings together.

💬

Two AI modes for different jobs

Fast-RAG handles quick citation-backed Q&A. Research-agent mode plans multiple steps, calls tools, reads evidence and keeps a visible trajectory for review.

Fast-RAG · quick cited answers
Research agent · multi-step, inspectable
🗂️

Structured evidence, not just uploaded PDFs

Papers are broken into chunks, claims, benchmarks, datasets and metrics with provenance, giving both humans and agents a shared evidence layer.

claim → source passage
benchmark → table, verified
📝

LaTeX writing IDE

Multi-file TeX editing, worker-side PDF builds, side-by-side preview, real-time collaboration, AI edit proposals and source zip export.

🧬

Research skill library

Reusable workflows for literature review, rebuttal prep, evidence audits, method comparison and paper packaging — shareable to the community.

🌐

Open community

Pick up community paper-project packs to help newcomers ramp up on the AI-assisted research process fast — and give your own work back.

🛡️

Team controls and BYOK

Per-project roles, viewer/editor boundaries, signed uploads, provider keys, model allowlists, routing priority, quotas and usage accounting.

⚙️

Durable async execution

Parsing, embeddings, searches, PDF builds and agent turns run as background jobs with queues, leases, retries, cancellation and live event timelines.

Build a project library.
Then write from it.

Create a workspace, invite collaborators, connect model providers, and keep the evidence trail close to the manuscript.