Academic Research Tools
Evidence-grounded academic literature discovery for AI agents, MCP clients, command-line workflows, and Python applications.
Academic Research Tools provides one normalized interface over several literature sources while preserving the source and identifiers of every record. Its core is independent of any agent harness; MCP, CLI, Python, and Hermes are adapters over the same implementation.
Providers
| Provider | Credential | Availability | Notes |
|---|---|---|---|
| arXiv | None | Core | Official public Atom API |
| Semantic Scholar | Optional SEMANTIC_SCHOLAR_API_KEY |
Core | Unauthenticated access uses shared public quota |
| Scopus | ELSEVIER_API_KEY |
Core, optional at runtime | Metadata and abstracts may require subscription or institutional IP |
| Google Scholar through SerpAPI | SERPAPI_API_KEY |
Experimental | Third-party SerpAPI integration; not direct scraping or an official Google Scholar API |
The package remains useful without credentials: arXiv is available immediately and Semantic Scholar can be queried without a key, subject to public rate limits.
Features
- Normalized
Paper,Author,SearchResult, andUnifiedSearchResultcontracts - Concurrent multi-provider search with partial-failure reporting
- Conservative deduplication by DOI, base arXiv ID, or normalized title plus year
- Per-record provenance and provider identifiers
- Local stdio MCP server for compatible clients
- Automation-friendly CLI and direct Python API
- Thin native Hermes Agent adapter
- Protected local credential configuration with environment overrides
- Interactive hidden-input credential wizard and offline diagnostics
- One-command Claude Code and Codex MCP plus skill installation
- Progressive onboarding that never blocks the first keyless search
- No telemetry
Installation
From GitHub
git clone https://github.com/istgrudd/academic-research-tools.git
cd academic-research-tools
python -m venv .venv
. .venv/bin/activate
python -m pip install .
From PyPI
pip install academic-research-tools
# or
pipx install academic-research-tools
No credential is required to verify the installation:
academic-research --version
academic-research doctor
academic-research status
academic-research search "traffic flow estimation low visibility" \
--sources arxiv,semantic_scholar --limit 5
Configure optional providers later, only when you need them:
academic-research configure
Secret input is hidden. Do not paste API keys into chat or pass them as command arguments.
Install the MCP server and workflow skill for a coding agent:
academic-research install --platform claude-code
# or
academic-research install --platform codex
The installer does not launch credential configuration. arXiv and Semantic Scholar public access remain ready for immediate use.
See Getting started and Credential setup.
MCP quick start
Run the local stdio server:
academic-research serve
If you use uvx without a prior install, configure the MCP client to execute:
uvx --from academic-research-tools academic-research serve
Generic MCP configuration:
{
"mcpServers": {
"academic-research": {
"command": "uvx",
"args": [
"--from",
"academic-research-tools",
"academic-research",
"serve"
],
"env": {
"SEMANTIC_SCHOLAR_API_KEY": "${SEMANTIC_SCHOLAR_API_KEY}",
"ELSEVIER_API_KEY": "${ELSEVIER_API_KEY}"
}
}
}
}
Only include environment variables for providers you intend to use. The protected user configuration is also read automatically. Details: generic MCP, Claude Code, Codex, Claude Desktop, and Cursor.
MCP tools
research_provider_statussearch_paperssearch_arxivsearch_semantic_scholarsearch_scopusget_scopus_abstractsearch_scopus_authorssearch_google_scholar
Python API
from academic_research import ResearchService
client = ResearchService.from_environment()
result = client.search(
"traffic flow estimation under low visibility",
sources=["arxiv", "semantic_scholar", "scopus"],
limit_per_source=10,
year="2020-2026",
)
for paper in result.papers:
print(paper.title, paper.doi, paper.provenance)
if result.errors:
print("Partial provider failures:", result.errors)
A provider that is not configured is excluded from ResearchService.from_environment(). The MCP unified search preserves results from available providers and reports missing optional providers with a configuration command. A provider-specific request returns an actionable error rather than silently substituting a different source.
Hermes Agent
Current Hermes installations can install the repository directly:
hermes plugins install istgrudd/academic-research-tools --enable
Pip-distributed discovery is also declared through the hermes_agent.plugins entry-point group. Scopus and SerpAPI use per-tool checks, so missing optional credentials never disable arXiv or Semantic Scholar.
See Hermes integration.
Credential safety
Credentials are resolved from environment variables first and a protected per-user JSON file second. The interactive wizard writes mode 0700 directories and mode 0600 files on Unix. The local file is plaintext and is not an encrypted vault.
This project does not:
- accept secrets as command-line arguments
- print credential values in
status - request credentials through an agent conversation
- display credential values in diagnostics or provider status
- store API responses automatically
- send telemetry
- require all providers to be configured
Do not commit .env; it is ignored by Git. See SECURITY.md.
Development
python -m venv .venv
. .venv/bin/activate
python -m pip install -e '.[dev]'
ruff check src tests __init__.py
pytest
python -m build
Tests use synthetic fixtures and do not spend provider quota. See CONTRIBUTING.md.
Scope and limitations
- Search ranking and coverage differ by provider.
- Citation counts from different providers are retained with provenance and should not be treated as directly interchangeable.
- Deduplication is intentionally conservative; ambiguous records may remain separate.
- This project does not bypass paywalls or grant access beyond the user's provider entitlement.
- Remote hosted MCP, operating-system keyring integration, and automatic literature-review decisions are outside the v0.2 scope.
Legal and data-provider notice
This project is an independent, unofficial integration and is not affiliated with or endorsed by Elsevier, Scopus, Semantic Scholar, arXiv, Google, or SerpAPI. Users are responsible for complying with each provider's API terms, acceptable-use policies, subscription conditions, and data licenses. The MIT license covers this project's source code, not provider data or services.
Thank you to arXiv for use of its open access interoperability.
License
Source code is available under the MIT License.
Release files for academic-research-tools 0.2.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| academic_research_tools-0.2.0.tar.gz | 62.2 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| academic_research_tools-0.2.0-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 104.5 kB
Release files / academic_research_tools-0.2.0.tar.gz
| Download URL | academic_research_tools-0.2.0.tar.gz |
|---|---|
| Size | 62.2 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
d43f2a7bece5d6540e7cf38de2007f9e7ae061db31d123e841411ff91cea8d26
|
|
BLAKE2b-256 checksum How to use checksums |
185cbcb678cc964153613b691f1ed8727d08541d8fe449e71fe467640ed57443
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 9, 2026.
Transparency logRelease files / academic_research_tools-0.2.0-py3-none-any.whl
| Download URL | academic_research_tools-0.2.0-py3-none-any.whl |
|---|---|
| Size | 42.3 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
86cbef3f40ee628402cc8c5fc5585ec579f88aa6f715a85d041182ab6f20f4da
|
|
BLAKE2b-256 checksum How to use checksums |
eafa11e83bc0e2dc78db6f32725854d6b86f475e0d58c3bc17ac06d6b4983dc9
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 9, 2026.
Transparency log