OSF CLI Go
# osf-cli-go
A Go command-line client for the Open Science Framework (OSF).
[](https://github.com/edithatogo/osf-cli-go/actions/workflows/ci.yml)
[](https://github.com/edithatogo/osf-cli-go/actions/workflows/lint.yml)
[](https://github.com/edithatogo/osf-cli-go/actions/workflows/security.yml)
[](https://pkg.go.dev/github.com/edithatogo/osf-cli-go)
[](LICENSE)
[](go.mod)
[](https://github.com/edithatogo/osf-cli-go/releases)
[](https://glama.ai/mcp/servers/edithatogo/osf-cli-go)
## Features
- `osf auth whoami` — Identify the authenticated OSF account
- `osf auth login` — Guided personal-access-token bootstrap for username/password users
- `osf projects list|get` — List and inspect projects
- `osf components list` — List project components
- `osf files list|download|upload|mkdir|rm` — Browse, download, upload, create folders, and delete OSF Storage files
- `osf search`, `osf search --bibtex`, `osf preprints list|search`, and `osf resolve` — Search/export OSF literature metadata, discover preprints, and resolve OSF DOI destinations
- `osf registrations create` — Create draft registrations for an existing node
- `osf export` — Export a node snapshot as JSON or a summary table
- `osf validate` — Check OSF node metadata with deterministic research-output or preregistration profiles
- `osf-mcp` — Stdio MCP server exposing read-only OSF tools for agent clients
- `osf completion bash|zsh|fish|powershell` — Shell completion scripts
- JSON and human-readable output modes
- Safe, atomic file downloads with conflict policy (fail/skip/overwrite)
## Install
Requirements:
- Go 1.26 or newer
```powershell
go install github.com/edithatogo/osf-cli-go/cmd/osf@latest
go install github.com/edithatogo/osf-cli-go/cmd/osf-mcp@latest
```
Or from a local checkout:
```powershell
go build -o bin\osf.exe ./cmd/osf
go build -o bin\osf-mcp.exe ./cmd/osf-mcp
.\scripts\build.ps1
```
## Authentication
Set `OSF_TOKEN` in your shell session. Do not commit the token or write it into project files. `OSF_USERNAME` and `OSF_PASSWORD` are supported as an opt-in fallback credential source, but personal access tokens remain preferred for automation and for accounts using SSO or two-factor authentication.
```powershell
$env:OSF_TOKEN = '<your-token>'
```
For guided token setup:
```powershell
osf auth login
```
## Quick Start
```powershell
osf --help
osf auth whoami
osf projects list
osf projects get https://osf.io/abc12/
osf components list abc12
osf files list abc12
osf files download --file <file-id> ./output/
osf files download --tree abc12 ./output/
osf files upload --node abc12 ./report.pdf
osf search "open science"
osf search "open science" --bibtex
osf preprints list
osf preprints search "open science" --provider osf
osf resolve 10.1234/example
osf registrations create abc12 --schema <schema-id> --title "Analysis plan"
osf export abc12 --json
```
## MCP Server
`osf-mcp` runs a stdio MCP server with read-only tools:
`osf_whoami`, `osf_projects_list`, `osf_project_get`,
`osf_components_list`, `osf_files_list`, `osf_contributors_list`,
`osf_search`, `osf_preprints_list`, `osf_preprints_search`, and
`osf_doi_resolve`.
The parity discovery tools are bounded and deterministic: `osf_search` requires
a non-empty query and accepts a limit from 1 to 100, while
`osf_preprints_list` accepts an optional provider and the same bounded limit.
`osf_preprints_search` requires a title query, accepts an optional provider, and
returns publication date, published state, DOI, and OSF HTML URL with a limit
from 1 to 100. These tools return structured JSON through the MCP server and
never perform writes.
OSF search results include stable IDs, title, abstract, keywords, year, and URL.
Use `osf search <query> --bibtex` to emit deterministic BibTeX records for
literature-review workflows. The command does not resolve contributor names or
download PDFs automatically.
For DataLad workflows, OSF CLI Go provides the general-purpose OSF API and
safe file primitives used around a dataset, including `osf export` and
`osf files download --tree`. It does not install or configure DataLad, Git
remotes, or git-annex special remotes; those integrations remain a separately
scoped interoperability boundary.
Local development configs are included for GitHub Copilot, VS Code, Claude,
Codex, Gemini CLI, and Qwen Code. Public registry metadata is in `server.json`
and `registry/`.
### Related MCP Servers
Other maintained servers in the same Glama portfolio:
- [FYI MCP](https://glama.ai/mcp/servers/edithatogo/fyi-cli) for freedom-of-information request workflows.
- [Healthpoint MCP](https://glama.ai/mcp/servers/edithatogo/healthpoint-rs) for licensed health-service directory data.
- [SourceRight](https://glama.ai/mcp/servers/edithatogo/sourceright) for reference and citation verification.
Related OSF ecosystem servers:
- [OSF connector by pipeworx-io](https://glama.ai/mcp/connectors/io.github.pipeworx-io/osf) for hosted OSF connectivity.
- [Paperclip](https://github.com/matsjfunke/paperclip) for multi-provider scholarly and OSF Preprints search.
## Output Modes
All commands support `--output table|json` and `--json` shorthand:
```powershell
osf projects list --json
osf auth whoami --output json
```
## Project Status
The CLI is **offline-tested** for read-only operations, file downloads, WaterButler write primitives, search, preprint listing, draft registration creation, project create/update/delete operations, and node export. All Conductor tracks are reconciled against their per-track plans with closeout review evidence; live OSF validation remains opt-in because it requires credentials and network access.
## Documentation
- [Release checklist](docs/release-checklist.md)
- [Documentation site source](mkdocs.yml)
- [Install guide](docs/install.md)
- [Usage guide](docs/usage.md)
- [Command reference](docs/commands.md)
- [Examples](docs/examples.md)
- [Architecture](docs/architecture.md)
- [Developer guide](docs/contributing.md)
- [MCP roadmap](docs/mcp-roadmap.md)
- [Contributing](CONTRIBUTING.md)
## License
Apache 2.0 — see [LICENSE](LICENSE).
## Citation
If you use this software in your research, please cite it using the metadata in [CITATION.cff](CITATION.cff).
TDQS
Scored across 6 tools
Each tool targets a distinct resource (components, contributors, files, project, projects list, user profile) with no overlapping purposes, making it easy for an agent to select the correct tool.
All tools share the 'osf_' prefix and use snake_case, generally following a 'noun_list' or 'noun_get' pattern. The exception is 'osf_whoami', which deviates slightly, but overall consistency is high.
With 6 tools, the server is well-scoped for basic OSF query operations. The count is appropriate for the focused read-only functionality offered.
The tool set covers read operations (list, get) for core OSF entities but lacks write, update, or delete actions, leaving obvious gaps for a full lifecycle. Agents can query but not modify data.