olegsotnikov/sallyport — explained in plain English
Analysis updated 2026-05-18
Let a coding agent make authenticated API calls without ever seeing the API key.
Approve or deny an AI agent's requests to run SSH commands on a server, one action at a time.
Keep an audit trail of every credentialed action an agent performed on your behalf.
| olegsotnikov/sallyport | facebook/quicklayout | shanggqm/codexu | |
|---|---|---|---|
| Stars | 329 | 348 | 295 |
| Language | Swift | Swift | Swift |
| Last pushed | — | 2026-07-18 | — |
| Maintenance | — | Active | — |
| Setup difficulty | easy | easy | easy |
| Complexity | 3/5 | 2/5 | 3/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
Requires an Apple Silicon Mac running macOS 14 or newer.
Sallyport is a free Mac app that keeps API keys and SSH credentials safe while still letting AI coding agents use them. Instead of handing a secret directly to an agent, which is risky because agents read config files and can be tricked into leaking data, Sallyport stores credentials in an encrypted local vault and performs the requested action itself. The agent asks for an operation over MCP, a protocol AI tools use to call external systems, Sallyport carries it out, and the credential never appears anywhere the agent can read it. The project was built in response to a real problem: coding agents and the packages they depend on regularly read secrets from environment files and shell variables, and recent supply chain attacks have specifically targeted those locations. Traditional secret managers still hand the actual secret to the workload running the task, which does not help when that workload itself cannot be fully trusted. Sallyport flips this by giving the agent an action to run rather than a secret to hold. Every action goes through an approval process. Depending on how a credential or connection is configured, it may require a one time approval using a click or Touch ID, or session level approval for new processes. All actions are recorded in a signed, tamper evident audit journal using the Mac's Secure Enclave hardware. The app supports HTTP requests, SSH command execution, and other configured tools reachable through MCP. Installation is done through Homebrew or a signed installer from the project's website, and setup to first use takes about two minutes. It requires an Apple Silicon Mac running macOS 14 or newer. The public repository holds source code for each release but does not accept outside contributions. It is released under the Apache 2.0 license, though the Sallyport name and logo are trademarked separately.
A Mac app that stores your API and SSH credentials in a vault and runs actions for AI agents without exposing the keys.
Mainly Swift. The stack also includes Swift, Go, MCP.
Use freely for any purpose, including commercial use, as long as you keep the copyright notice and give credit for changes.
Setup difficulty is rated easy, with roughly 5min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
double-check against the repo, no cap.