OpenDraft

Turn one topic line into a research draft whose every DOI was confirmed by at least two scholarly databases.

Stars
★ 537
Last updated
17d ago
License
MIT
Primary language
Python

At a glance

How it runs
CLILibrary / SDKAgent plugin / skill
Works with
Universal · cross-platformCodex · Claude Code · OpenAI API · Claude API
Cost
Free software; you pay for model usage
Setup effort
Medium · a few setup steps
You'll need
Python 3.10+Google Gemini API keyOpenAI API key (optional)Anthropic API key (optional)ElevenLabs API key (optional, for audio digests)Shell / CLINetwork accessLocal filesystem
Typical use
A graduate student starting a master's thesis or PhD dissertation who wants a 30-80 page structured first draft with a real, checkable bibliography to rewrite from.
Not a fit if
  • Writers who want a submission-ready final paper with no human review
  • Users who only need light text polishing and no citation checking
Website
opendraft.xyz ↗ set on GitHub
Maintenance
Commits in the last 30 dayslast commit 2026-11-04latest release v1.8.1 (2026-09-29)+0.7 ★/day over 7 days

What does this agent do, and when should you use it?

OpenDraft is an open-source Python engine that turns a single topic line into a first draft of a research paper, literature review or thesis chapter, typically 10,000-20,000+ words and 5-80+ pages in 10-20 minutes. It runs 19 specialized agents across six phases: research, structure, writing, citation, polish and export. Its differentiator is citation handling: candidates are found via CrossRef, OpenAlex and Semantic Scholar, and by default a citation survives only if its DOI is held by at least two of those three databases; single-source and unconfirmed results are dropped and logged. Every entry in bibliography.json carries provenance fields such as verification_status and verification_sources, and a separate CitationClaimVerifier judges whether a source plausibly supports the claim it is attached to. You can run it three ways: as a CLI, as a Python library imported into your own code, or as the autonomous-research Skill for Claude Code and Codex. The software is MIT-licensed and free, but you bring your own model API keys, at roughly $0.35-$3.00 of API cost per draft.

The pipeline orchestrates 19 agents through RESEARCH, STRUCTURE, WRITING, CITATION, POLISH and EXPORT phases. Research queries the CrossRef, OpenAlex and Semantic Scholar APIs for candidates, then looks each DOI up directly; with the defaults require_multi_source=True and min_confirming_sources=2, anything confirmed by fewer than two databases is dropped. The citation phase dedupes, quality-filters, and runs CitationClaimVerifier against the paper topic, returning RELEVANT, IRRELEVANT or UNCERTAIN and writing citation_claim_verification.md plus .json (irrelevant citations are removed only above CLAIM_VERIFICATION_MIN_CONFIDENCE, default 0.7). Output exports to PDF, Word (.docx) or LaTeX via to_pdf(), to_docx(), to_latex(), in 57+ languages. Companion commands include opendraft tldr (five-bullet summary), opendraft digest (ElevenLabs audio briefing), opendraft revise (versioned AI revision), opendraft data (World Bank, Eurostat, Our World in Data), and --expose for a fast scoping report.

  1. A graduate student starting a master's thesis or PhD dissertation who wants a 30-80 page structured first draft with a real, checkable bibliography to rewrite from.
  2. A researcher scoping a literature review who needs to confirm enough verifiable literature exists before committing (run --expose for a three-times-faster overview).
  3. Someone preparing a journal submission who needs LaTeX source at the end for the publisher's template.
  4. A lab that wants a DOI-bearing candidate source list with per-citation provenance before running its own formal search.
  5. A Claude Code or Codex user who wants the research workflow as an editable Skill rather than a service call, with no API key of its own.
  6. A reader who has someone else's PDF paper and just wants opendraft tldr or opendraft digest to get the gist or a 60-second audio briefing.

Key setup facts

Pulled from the install and usage notes below. The project README has the current values.

Install / run
  • pip install -r requirements.txt
  • pip install opendraft[pdf]
  • npx skills add federicodeponte/opendraft --skill autonomous-research
Environment variables
GOOGLE_API_KEY OPENAI_API_KEY ANTHROPIC_API_KEY ELEVENLABS_API_KEY
Config files
.env
Runtime
Python 3.10+

How do you install or deploy this agent?

Prerequisites: Python 3.10+ and a free Gemini API key.

git clone https://github.com/federicodeponte/opendraft.git
cd opendraft
pip install -r requirements.txt

Create a .env in that directory with your key (Gemini is the default provider):

GOOGLE_API_KEY=your-gemini-api-key

PDF reading in the TL;DR and digest tools needs an optional extra:

pip install opendraft[pdf]

If you only want the agent Skill (no Python engine), install it directly:

npx skills add federicodeponte/opendraft --skill autonomous-research

Other providers need extra configuration: OpenAI uses AI_PROVIDER=openai plus OPENAI_API_KEY, Anthropic uses AI_PROVIDER=claude plus ANTHROPIC_API_KEY. Audio digests additionally require ELEVENLABS_API_KEY.

How do you use this agent?

As a Python library:

from engine.draft_generator import DraftGenerator

generator = DraftGenerator()
draft = generator.generate(
    topic="The Impact of AI on Academic Research",
    paper_type="master",  # research_paper, bachelor, master, phd
    language="en"
)

draft.to_pdf("thesis.pdf")
draft.to_docx("thesis.docx")
draft.to_latex("thesis.tex")

From the command line:

# Fast scoping overview instead of a full draft
opendraft "Neural Networks in Healthcare" --expose

# Revise an existing draft; writes draft_v2.md plus PDF/DOCX
opendraft revise ./output "Make the introduction longer and add more context"

# Five-bullet summary of any paper
opendraft tldr paper.pdf -o summary.md

# 60-second audio briefing (needs ELEVENLABS_API_KEY)
opendraft digest paper.pdf --voice adam

In an agent, after connecting Edge, the invocation is just Write a paper on <topic>. That Skill is 18 markdown prompt stages plus 5 standard-library Python scripts, does not call this Python engine or any hosted service, and needs no API key because the agent reading it is the model.

What are this agent's strengths and limitations?

Pros
  • Multi-source DOI confirmation is the default, not an opt-in: fewer candidates get through, but each surviving citation is held by at least two databases, and single-source drops are logged.
  • Provenance is unusually explicit - verification_status and verification_sources are written even when empty, so an unconfirmed citation cannot serialize to look confirmed.
  • A second, separate check (CitationClaimVerifier) targets whether a source supports the claim it is cited for, and the README states plainly that these are LLM judgements, not proofs.
  • Three documented entry points - CLI, importable Python library, and an Apache-2.0 Claude Code / Codex Skill whose 18 stages and 5 scripts are files you can open and edit.
  • MIT licensed with a real export surface (PDF, Word, LaTeX), 57+ languages, and swappable model providers (Gemini default, OpenAI, Anthropic).
Limitations
  • Free software but bring-your-own keys: roughly $0.35-$3.00 per draft, plus a hard network dependency on scholarly APIs and a model provider.
  • Strict confirmation shrinks the bibliography enough that a run can now fail outright with PipelineValidationError when zero citations survive, and the quoted citation-count and page-length figures predate multi-source confirmation and have not been re-measured.
  • The three databases are not independent: OpenAlex and Semantic Scholar both ingest Crossref metadata, and one of the confirming two may be the database that supplied the candidate in the first place, taken at its word.
  • No arXiv API client, and claim-level checking runs against the paper topic during the citation phase because no draft text exists yet - sentence-level checking requires a separate call to run_citation_claim_verification().
  • Explicitly not a one-click author: human review, your own analysis, and compliance with your institution's AI policy are required, and the output is a draft, not a submission-ready paper.

How does this agent compare with similar options?

The README positions OpenDraft against general-purpose AI writing tools that produce confident prose with hallucinated or unverifiable citations, and argues for a multi-agent pipeline grounded in real literature instead. It also points to related surfaces: the hosted OpenPaper app at openpaper.dev, the Edge skill gateway at getedge.cc, and ChatGPT-style single-model alternatives that the project's chatgpt-alternative topic tag names.

Key facts side by side with the most closely related agents.

Agent Source review Form / cost Stars Updated Language Full support on
OpenDraft This agent 54 · Major gaps CLIFree + model costs ★ 537 17d ago Python Codex · Claude Code · OpenAI API · Claude API
Research Agent Skills 68 · Some gaps CLIFree ★ 174 4d ago Python Codex · Claude Code
PaperJury 72 · Some gaps Agent plugin / skillFree + model costs ★ 1.2k 3mo ago JavaScript Codex · Claude Code
PaperDebugger 35 · Major gaps Browser extensionFree + model costs ★ 1.6k 4mo ago TypeScript —

How does FollowAgents rate this agent?

FollowAgents source review · FARS-2.1
Major gaps
54/ 100 5-point scale 2.7 / 5
Trust 12/29
Reliability 6/14
Adaptability 12/18
Convention 12/18
Effectiveness 7/13
Verifiability 5/8
Why each dimension lost points
Trust12 / 29 · 2.1/5

The README says users provide their own model key and names Gemini, Claude, OpenAI, and bibliographic metadata services; the CLI stores configuration in the user directory. Security guidance says not to commit keys and recommends checking citations independently, providing some data-flow and key-handling context. Deductions: the packet does not explain key protection, transmission or retention of inputs and generated files, the scope of network access, or adequate confirmation, revocation, and recovery procedures.

Reliability6 / 14 · 2.1/5

The README describes a multi-phase process and a rule requiring a DOI to appear in at least two metadata sources. Requirements and the CLI provide guidance for common key, network, quota, and export errors. Deductions: the README says 19 agents while the skill says 18 stages plus one evidence stage; dependencies have version ranges but no lockfile or availability guarantee is shown, and error handling covers common cases rather than all failures.

Adaptability12 / 18 · 3.3/5

The README and skill identify research papers, literature reviews, and thesis drafts as intended uses, and list degree levels, language options, and skill triggers. The skill asks for venue, word limit, citation style, and document type, with a stated default path if the user declines. Deductions: model and environment options are described, but the packet does not fully show all supported scenarios, boundaries, or configuration limits.

Convention12 / 18 · 3.3/5

The repository includes a README, documentation links, quick install and run steps, sample output, FAQ, contribution guidance, security policy, package version metadata, a changelog link, and an explicit MIT license. Deductions: some information is only linked externally and absent from this packet; OpenDraft, OpenPaper, and the skill name are used together, creating some naming ambiguity; maintenance responsibility has a vulnerability-reporting path and response promise but no specific maintainer update plan.

Effectiveness7 / 13 · 2.7/5

The materials describe citation-bearing research drafts and PDF, Word, and LaTeX exports; the CLI also offers revision, summaries, and outline/source-only output. The README explicitly says drafts require human review and gives an estimated cost of $0.35–$3 per draft. Deductions: utility and cost are mostly project claims, with insufficient evidence here to establish quality or turnaround across topics.

Verifiability5 / 8 · 3.1/5

The README describes multi-source DOI checks and distinguishes a work's existence from whether it supports a particular claim; the skill also describes citation-compilation scripts and an evidence stage. Deductions: these methods and outcomes are primarily asserted in project files; the packet does not show verification results, item-level source mapping, or independent evaluation, so higher scores are not supported.

Risks and how to mitigate them
  • Edit generated drafts and follow institutional disclosure rules; a DOI's existence does not show that its source supports a particular claim.
  • Understand what inputs are sent to model providers and external metadata services; protect API keys and do not commit them to the repository.
Evidence confidence: Low Reviewed Nov 08, 2026 Reviewed revision 39b21f74b3f5
See the full review method →
Maintain this project? Add the FARS badge to your README

The badge shows the current review score and verdict and links back to this page. It refreshes about once an hour after a re-review.

FARS

[![FARS review](https://followagents.com/badge/opendraft.svg)](https://followagents.com/en/agents/opendraft)
<a href="https://followagents.com/en/agents/opendraft"><img src="https://followagents.com/badge/opendraft.svg" alt="FARS review"></a>

Recent releases

Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.

  1. v1.8.1
    Release notes excerpt
    Fixes since 1.8.0: - #31: user-supplied sources and context now reach the paper; minimum citation density scaled to length. - #32: working Claude provider (AI_PROVIDER=claude), configurable models and API base URLs. - #84: dependency bounds raised (openai 2.x, lxml 6, rich 14, and more). - README hero refresh.
    View on GitHub ↗
  2. v1.8.0 v1.8.0 — Multi-source citation confirmation
    Release notes excerpt
    Multi-source citation confirmation OpenDraft now keeps a citation only when the source can be independently confirmed, closing the gap between a fallback chain and real verification. What's new - Multi-source DOI confirmation (default on). A DOI-bearing citation is kept only if 2 of {Crossref, OpenAlex, Semantic Scholar} independently hold that exact DOI. Single-source results are dropped and the drop is logged, never silent. …
    View on GitHub ↗
  3. v1.7.4 v1.7.4 — LaTeX (.tex) export
    Release notes excerpt
    Added - LaTeX (.tex) output alongside PDF, DOCX, and Markdown. Every generated paper (and research exposé) now also produces a standalone, compilable <name>.tex next to <name>.pdf/<name>.docx, bundled into the ZIP. The .tex reuses the same Pandoc pipeline and preamble as the PDF path, so it carries the same style-formatted References/bibliography (APA/IEEE/Chicago/MLA) and compiles with XeLaTeX (ships a % !TeX program = xelatex header). If Pandoc is missing the .tex step is skipped with a warning; the run still completes with its PDF/DOCX. Fixed …
    View on GitHub ↗

FAQ

Does it hallucinate references?
By default a citation is kept only if its DOI is held by at least two of CrossRef, OpenAlex and Semantic Scholar; single-source and unconfirmed results are dropped, and the LLM-asserted fallback (enable_llm_fallback) is off by default and permanently tagged llm_unverified when enabled. Be precise about what that buys you: it proves the work is registered and indexed, not that it supports the sentence it is attached to.
What does a draft cost to run?
The software is MIT-licensed and free; you pay only your own model API usage. The README quotes roughly $0.35 per draft on Gemini Flash and up to about $3.00 on Claude Opus. Audio digests are the one extra dependency, requiring an ElevenLabs account.
Why did my run fail with PipelineValidationError instead of producing a draft?
Strict multi-source confirmation, the strict quality filter and claim-level removal all shrink the bibliography, and the pipeline raises that error when no citations survive the citation phase. The documented fix is to widen the search or deliberately relax settings such as require_multi_source=False, understanding that citations then carry verification_status: not_checked.
Is the Skill the same codebase as the Python engine?
No - it is a port, not a wrapper. The Skill queries Crossref, OpenAlex and DataCite only (no Semantic Scholar client), so multi-source confirmation is not its rule; instead every DOI it prints must resolve at Crossref or DataCite, and citations.py compile refuses to render a bibliography containing one that did not, naming the offender and exiting nonzero. Export goes through pandoc rather than this repository's PDF pipeline, and the bundle is Apache-2.0 with MIT notices retained in THIRD_PARTY_NOTICES.md.
Can I use it commercially?
Yes. The MIT license permits commercial use, modification and distribution without restriction, and the README invites building products and services on top of it.
Is OpenDraft open source?
Yes. The source code is on GitHub (federicodeponte/opendraft) under the MIT licence.
Is OpenDraft safe to use?
FollowAgents' FARS review (Nov 08, 2026) scored OpenDraft 12/29 on trust and safety. None of the eight safety controls the review checks for was missing outright.
Is OpenDraft still maintained?
Yes. The latest commit on GitHub is from 2026-11-04, within the last 30 days. The latest release is v1.8.1 (2026-09-29).
Who makes OpenDraft?
OpenDraft is published on GitHub by the account federicodeponte. FollowAgents has not verified which person or company is behind that account.
What is OpenDraft's official website?
The project's GitHub repository lists https://opendraft.xyz as its website. That link is set by the repository owner; the source code itself is on GitHub (federicodeponte/opendraft).
Was this page helpful?
View on GitHub ↗ Install ↓

Compare agents like this one

The same FARS review applied across the shortlist this agent qualifies for.

Related concepts