Nullgap attacks your code and running apps in a sealed sandbox, reports only what it can exploit, and re-runs the same exploit against the fix. No proof, no finding.
Free 14-day trial · Sign in with GitHub · or talk to a security engineer
Maps how your data is handled to 31 standards and laws
Where developers already work
Can you spot the vulnerability?
Click the line you think is unsafe. Then see how Nullgap proves it, instead of just pointing at it.
Hint: who is allowed to see invoice 1002?
Not that one. Look at what the database is asked for.
You found it. Any signed-in user can read any invoice. Here’s the proof Nullgap attaches.
GET /invoices/1002 signed in as customer A · invoice owned by B → 200 · B’s name, address and amount returned Worked 2/2 control: A opens A’s own invoice → normal No effect after the fix: same request → 404 BlockedFind these in your code
Stop triaging maybes.
Most tools hand you a list of things that might be wrong. Nullgap hands you the few that are, with the proof.
Nullgap reads your code and your running app the way an attacker would, and lists what looks exploitable.
Each lead is attacked in a sealed sandbox. Only what really works is reported, with the requests that prove it.
Nullgap writes the fix, re-tests it on a fresh checkout and opens a pull request for your team to review.
The same exploit runs again against the fix. It has to fail before the finding is closed.
A finding is a claim. A proof is four checks that all pass.
Nullgap marks a finding proven only when the attack works, works again, a harmless control doesn’t, and a separate review agrees. A status code alone never counts. Anyone on your team can replay it.
| Run | What it showed | Result |
|---|---|---|
| Attack run | Customer A fetched customer B’s order export | Worked |
| Control run | An ordinary, harmless input behaved safely | No effect |
| Repeat | A second run gave the same result | Worked |
| Independent check | A separate reviewer agreed it shows the flaw | Accepted |
One proof standard, from the pull request to the running app.
Every pull request reviewed for security before it merges, with the proof attached when a flaw can be shown. Triage by replying in the PR.
A full test of a repository from its code. Each finding is reproduced in a sealed sandbox you can replay. Price shown before it runs.
Point it at staging with up to five test accounts. The access matrix proves when one user can reach another user’s data.
Thirty-two prompt-injection attacks against your chatbot or agent, caught when it leaks a secret or calls a tool it must not.
Malicious and risky packages flagged during a pentest, with package rules you can enforce.
A fix pull request for each proven finding, re-tested against the original exploit before it is called fixed.
Ask a Nullgap security researcher to check a finding or a whole report by hand, with notes on the finding.
Ask Nullgap about findings in Slack, and see proven findings and apply verified fixes in VS Code.
The same four states in pull request comments, Slack threads, Jira issues and the editor.
Built for the security review your security team will run on us.
Every point below is something you can check in the product, the docs or your own audit log.
Four checks must pass, so your team stops triaging false positives.
The exact requests and sandbox output ship with each finding.
The original exploit is re-run against each fix before it closes.
Whitebox pentest, PR review and live testing under one standard.
Up to five identities per app, with a proven access matrix.
Prompt-injection testing mapped to the OWASP Top 10 for LLM apps.
Priced by code size, one free retest, refunds when a run fails on our side.
Expert review by a Nullgap security researcher on request.
GitHub, Slack, Jira, VS Code, MCP, API and webhooks.
| AI prompts and responses are never stored | AI processing runs only with providers under zero data retention. If that can’t be confirmed, the run stops. |
|---|---|
| Credentials are never sent to the AI or stored | Secrets found in your code are masked before storage and never sent for analysis. |
| Your source code is not kept | A repository is checked out for a run and deleted when the run ends. |
| Exploits run in a sealed sandbox | Proofs run in an isolated environment with no outside network. |
| Live tests only reach what you approve | Every request is checked against your scope list, and private addresses are never reached. |
| Test accounts are sealed | Passwords and tokens are encrypted for your workspace and never shown again. |
| Automatic deletion | Set a retention period and stored data is removed when it ends. |
| Export and erase on request | An owner can download everything stored for the workspace, and erase it, at any time. |
| Tamper-evident audit log | Every change is recorded with who made it and when, in a hash-chained log. |
| Access control | Owner, admin and member roles, SSO and SCIM, required MFA, and a fresh code for sensitive actions. |
| Support access is opt-in and audited | Nobody at Nullgap reads your data unless an admin grants it for a set time, and every read is logged. |
| Incident notification | If an incident affects your data, the contact you set is told without undue delay. |
Pick the frameworks you answer to, and Nullgap shows how its handling of your data lines up with each one, control by control.
This is a mapping of how Nullgap handles your data, not legal advice or a certification. Using Nullgap does not by itself make your organisation compliant.
- Pentest attestation letter
- Executive summary and full report
- Evidence bundle for every finding
- Audit log export
- Findings export as CSV or SARIF
- SOC 1 Type II In process
- SOC 2 Type II In process
- Data processing agreement with sub-processor list
- Security questionnaire answers on request
- Responsible disclosure policy
Priced in the open. You see the price before anything runs.
And the proof guarantee: if a pentest proves nothing, your next pentest of that repository is half price.
- 14-day trial of PR review
- Watch a pentest find and prove
- No card to start
- Every pull request reviewed, proof attached
- Triage from the PR, Slack and VS Code
- 20 fix credits per developer each month
- Proven findings and verified fixes
- Live testing with up to 5 accounts
- Free retest within 60 days
- PR review, 4 pentests and AI app testing
- Expert review of every report
- SSO, SCIM, audit export, DPA, named engineer
| Expert review by a Nullgap researcher | $250 / finding | Or $2,500 for a whole pentest report. Included on Enterprise. |
|---|---|---|
| AI app test | Credits | Prompt-injection run against one AI app; the most it can use is shown before it starts. |
| Fix runs beyond the included credits | $1.50 / credit | Charged only for work done; refunded if a run fails on our side. |
| Open-source projects | Free | PR review on public open-source repositories. |
See Nullgap prove a vulnerability in your stack.
A 30-minute call with an engineer, not a sales script. Bring a repository or a staging app and we’ll walk through a live proof.
- Scoping for pentests and live testing
- Security review and questionnaire answers
- Enterprise terms: SSO, audit log, expert review
Questions security teams ask
Is my source code stored?
No. A repository is checked out for a run and deleted when the run ends. Findings keep the line they refer to and the explanation.
Is my code used to train AI models?
AI processing runs only with providers under zero data retention, so prompts and responses are not stored. If that can’t be confirmed, the run stops.
Will you attack production?
Proofs run in a sealed sandbox built from your code. Live testing only reaches addresses you put on your scope list, read-only unless you allow more.
What happens when Nullgap can’t prove something?
It isn’t reported as a finding. You can still see what was looked at and ruled out, so you know what was covered.
How is a pentest priced?
By the size of the code: $750 up to 50k lines, $2,500 up to 250k, $6,000 up to 1M, and a quote above that. You see the exact price before it runs, a retest within 60 days is free, a run that fails on our side is refunded, and if a pentest proves nothing your next one of that repository is half price.
Where is Nullgap on SOC 1 and SOC 2?
Our SOC 1 Type II and SOC 2 Type II audits are in process. Ask for the security pack for our current controls, data processing agreement and questionnaire answers.
Can we use it for our audits?
Yes. Pentests produce an attestation letter, an executive summary and a full report, and the compliance mapping shows how data is handled for the frameworks you choose.
Get your first proof today.
Connect a repository and watch Nullgap find, prove and fix. You only ever see what it could prove.