Alerts
When something needs you, Scopebond tells you by email, in Slack or through a webhook.
Product
See what your agents do. Block the risky ones. Prove it.
One workspace for everything your AI agents do, the actions you block and the records that back it up.
See what they do
Your workspace, tab by tab.
What needs you, everything your agents did, and how each rule is set.
Overview
What needs you
- Findings and unblock requests, in one list
- Every computer, protected or not
- What was blocked, ready to share
Activity
Everything your agents did
- Allowed, monitored and blocked, side by side
- Search and filter by computer, person, repo or tool
- Package installs and services used, listed too
Protection
Your rules, your pace
- Monitor or Block, rule by rule
- Protection packs update themselves
- Safe Install on every package an agent installs
Block what matters
Protection where a mistake costs the most.
Scopebond covers the places an agent can do real harm, and blocks the actions you choose.
Code and branches
Force-pushes, protected branches and release settings.
Secrets
Passwords, keys and the files that hold them.
Cloud and infrastructure
Deleting resources, changing access and production deploys.
Databases
Resets, dropped tables and risky migrations.
Packages Safe Install
Known malware and made-up or look-alike names are blocked; brand-new packages ask first. Works offline.
AI tool connections
The MCP servers and services your agents call.
Protection packs, kept up to date for you
Packs come from Scopebond Cloud and update themselves. When we add protection for a new tool or a new threat, every computer gets it.
- Updates arrive on their own
- Signed by Scopebond
- The same protections on every computer
Protection packsUp to date
- Code and branchesCurrent
- SecretsCurrent
- Cloud and infrastructureCurrent
- DatabasesCurrent
- PackagesCurrent
- AI tool connectionsCurrent
Roll out safely
Monitor first. Block when ready.
Nobody is held up on day one, and you decide when each rule starts blocking.
- Monitor
See what would be blocked
Every rule starts in Monitor. It records what it would have blocked, and your developers keep working.
- Block
Switch on, one rule at a time
Flip a rule to Block once its Monitor results look right. Removing Scopebond from a computer alerts you.
- Unblock
Unblock in seconds
A developer asks. An admin or the agent's owner approves, once or for a while, and the record shows who approved it.
Prove it
The answer is already signed.
When a client, an auditor or your own team asks what an agent did, you have a signed record of each decision and reports that sum it up.
Monthly report
What your agents did and what was blocked, summed up each month and ready to forward.
Evidence packs
On the Audit plan, export the signed records a client or auditor asks for as one evidence pack.
Every decision is signed.
The computer that decided signs the record, so a client or auditor can check it hasn't been changed.
Works with
The AI tools your team already uses.
Set up once per computer. Your rules follow every agent.
Early access: a person has tested Claude Code on Windows 11. The other tools, and macOS and Linux, pass the automated tests and are in beta. What has been tested.
Pull requests
The Scopebond GitHub Action runs as a required check. An agent's pull request that breaks your rules is blocked from merging, with a signed record for each one.
Cloud agents
Claude Code on the web, Codex cloud and the Copilot coding agent, through a repository setting.
Need something else?
Tell us what your team needs. We reply by email.
See it with sample data.
The sample workspace needs no account. Free for one agent; see pricing.