Aikido

What is Aikido?

Aikido is an application security platform combining code, cloud and runtime protection in a central system.

AI code review, source analysis and AI pentesting address different stages of development, with separate access and charging conditions.

Core features

Code scanning and contextual review

• Dependency security: Software Composition Analysis checks known CVEs and risky open-source licenses. It runs on pushes and pull requests, in the IDE and through scheduled rescans, daily by default.
• Static analysis: SAST checks source for known dangerous patterns without needing a running application. Logic or authorization flaws that do not resemble those patterns can remain outside its coverage.
• AI code review: Deep Review explores relevant code beyond the diff, considers the wider impact of a pull request and posts feedback on that request. It can flag a missing access check or broken pagination caused by the change, covering security and quality.

AutoFix and dependency updates

AutoFix proposes reviewable changes for dependencies, application code, infrastructure as code, containers and pentest findings through inline IDE diffs, CI/PR suggestions and manually or daily generated pull requests.

Dependency AutoFix prefers the minimum fixed version, usually a patch or minor upgrade; a major upgrade means no lower version fixes the issue.

Cloud visibility and runtime controls

Cloud Discovery searches resources, risks and misconfigurations across accounts; Cloud Inventory maps assets, and custom CSPM rules flag risky configurations.

Zen Firewall embeds directly in application code and runs on the same server as the application. Its listed controls cover dangerous user input, injection and path-traversal attacks, route and user rate limits, malicious traffic such as bots or known attackers, country-based access and outbound traffic monitoring.

Guide

Repository and pull-request setup

1. Create the account through a supported source-control connection and connect the repositories to scan. Connecting repositories starts six scan types: SCA, SAST, Secrets, Licenses, IaC and Malware.
2. Enable SAST and Code Quality in PR Checks for the relevant repositories so those checks inspect newly added code.
3. Enable Deep Review separately in PR Checks for contextual security and quality review beyond the diff.

Use cases

Authorization changes in a pull request

Aikido gives a PR exposing one user's orders to another as an example for Deep Review: the app still runs and SAST may see no known dangerous pattern.

Customer isolation in established software

AI Code Analysis examines existing source when a team is unsure whether an established application isolates customers from each other.

Internet exposure and forgotten assets

Attack Surface Monitoring discovers reachable assets such as an old login page or an admin API exposed by a proxy change. Quick Scan provides ongoing exposure checks on already-known domains.

Who is it for

• Security teams working with developers: Aikido positions its platform around enforcing controls across the SDLC and using agents to support and educate developers.
• Teams keeping source on their own machines: Local scanning supports development checks, isolated environments and sensitive local repositories, with a compatible workspace required. Aikido says standard integrations are faster and provide better coverage for most workflows.
• Organizations with distinct pentest roles: Manage Pentests controls who can create, run and configure tests. Admins have it, default users may be granted it, and team-only users cannot receive it. Default users can access workspace pentests; team-only users see a pentest only when they can access every repository attached to it. Without management permission, users can still open accessible pentests, work on findings and retest deployed fixes.

Platforms

Source control, editors and CI

Aikido connects through GitHub, GitLab, Bitbucket, Azure DevOps and on-premise source-control setups. The available IDE routes include Visual Studio, VS Code, JetBrains IDEs, Cursor, Windsurf, Kiro, Google Antigravity and Eclipse.

Local scanner documentation includes CI/CD setup for Azure DevOps Server, Bitbucket Pipelines, Bamboo, CircleCI, GitHub Actions, GitLab Self-Managed, Jenkins and TeamCity, alongside Linux, Mac and Windows setup guides.

MCP setup routes are published for Claude Code, Cursor, Codex CLI, Gemini CLI, JetBrains AI, GitHub Copilot, Mistral Vibe, OpenCode and Devin. API Authorization and Rate Limiting documentation is also available.

Scanner and runtime language boundaries

SAST distinguishes cross-file and within-file taint tracing. JavaScript, TypeScript, PHP, .NET/C#, Java, Rust, Go, Ruby, Python, Scala, C/C++ and Visual Basic are listed for cross-file analysis. Swift, Android, Kotlin, Dart, Elixir, Apex and Clojure are listed for within-file analysis.

Zen's supported runtimes are Node.js, Python, PHP, Java, .NET Core and Framework, Ruby, and Golang, with Golang marked beta.

Pricing

Platform subscriptions

USD prices captured on October 28, 2026 use 10 users per paid plan; annual prices are displayed yearly totals. Prices exclude withholding, import, sales, VAT and other local taxes.

Developer requires no credit card and lists rescans every three days. Plan-card quotas are subject to fair use. A dash means the card lists no VM scaling-group allowance.

Aikido Max is offered by quote as an add-on with unlimited monthly credits for all AI features, including Deep Review, subject to fair use.

Wallet balance and spending limits

One Aikido credit equals $1 USD. Eligible active Pro and Advanced workspaces can receive 100 and 200 complimentary credits per calendar month respectively; Developer and Basic do not include these top-ups. Purchased and complimentary credits share a non-expiring wallet balance. No wallet use may mean no complimentary top-up that month; the existing balance remains.

Credit purchases have a 500-credit minimum per transaction. Payment can use a card or an invoice, but invoicing is available only to paying customers, with credits added on successful payment.

Workspace admins can set per-feature spending caps, with reset periods depending on the feature. On reaching a cap, Aikido skips actions that would charge credits until the limit resets.

Deep Review costs one credit per pull request, with no extra charge for additional reviews on that same PR. Reviews are skipped when credits run out. Its monthly cap skips chargeable reviews until the next calendar month.

Pentest scope and usage charges

A Standard Pentest costs 4000 credits for a time-boxed assessment of one application and its primary APIs. Rightsized Pentests use connected repositories to map attack vectors and determine security-relevant testing scope, with the total confirmed before testing starts. Rightsized displays $50–$30,000+, scoped from repositories, endpoints and roles.

Continuous Pentesting is usage-based at 10 credits per agent launched. Its annualized estimate uses the last 30 days of Git activity. Aikido checks configured deployments or scheduled runs but launches agents only for security-relevant changes, so actual charges depend on launches rather than the estimate alone. Initial-pentest findings can be retested for up to six months, including checking for bypasses of the bug fixes.

Run now, pay later

The first full assessment can begin using pending credits without an upfront credit purchase. High and Critical findings remain blurred until payment; larger assessments reveal at least one such finding as a preview. The report and retesting become available after the pending balance is paid or waived. Any outstanding balance must be paid before using this offer again.

Eligibility requires no unpaid credits, a pentest costing at most 8000 credits, at least one active non-demo repository, at least two valid test accounts, and a work email either for the user or a workspace admin. Personal email addresses do not qualify.

Waiver requires all of these conditions: the workspace's first full assessment, use of Run now, pay later, a Rightsized Pentest at recommended coverage or deeper, correct setup permitting thorough testing, and completion without High or Critical findings. The no-pay condition excludes below-recommended coverage, second runs and retests, Continuous Pentesting, narrow feature or endpoint tests, and incomplete setup such as missing users or inadequate application context. It can also be withheld for a Medium or Low finding when Exploit Further was available but unused and the finding could reasonably have escalated to High or Critical severity. The condition concerns severity, including more than privilege escalation.

Renewal and termination

The public MSA provides automatic renewal for a period equal to the current term, with written notice at least 30 calendar days before term end to terminate at that term's end. If termination is for a reason other than Aikido's uncured material breach, the customer remains responsible for fees covering the rest of the term. Termination for Aikido's uncured material breach carries a pro-rata refund of prepaid unused fees.

The free card says “free forever,” while the MSA gives the free version no minimum or guaranteed term and permits Aikido to terminate it at any time for any reason.

Alternatives

Snyk lists code, open-source dependency, container, infrastructure-as-code, API/web DAST and secrets products: a comparison based on developer-scanner scope.

SonarQube covers code verification, quality and security; SonarQube for IDE provides on-the-fly analysis and coding guidance. SonarQube Cloud is fully managed SaaS; SonarQube Server is self-managed inside the customer's perimeter.

Limitations

Detection coverage and operational risk

• Uncertain reachability: Reflection, dynamic imports, metaprogramming and complex build steps can obscure parts of the call graph. Aikido describes a conservative approach: an ambiguous finding is retained or has its severity lowered rather than being suppressed as unreachable.
• Known-domain checks: Quick Scan covers exposure and configuration such as headers, cookies and TLS. It is not the tool for deep injection testing or proving whether users can access one another's data.
• Pentest scope: Aikido Pentest excludes social-engineering and physical-security assessments and does not guarantee that every vulnerability is found.
• Target effects: The pentest addendum describes autonomous agents as potentially unpredictable and non-deterministic, with risks of degraded performance, service interruption or data alteration. Customers configure the targets and context. Results are point-in-time and may include false positives, omissions, or AI observations that cannot be reproduced or fully explained.

Authorization and credential testing

Targets must be explicitly designated and owned by the customer or lawfully authorized for testing. Third-party targets require that party's explicit permission. Application tests default to non-production, staging or test environments without live user data or risk to live business operations, unless both parties expressly agree otherwise in writing. Network Pentests may include production targets expressly selected in scope, with acknowledgment of the heightened service-impact risk.

Credential-Based Testing is optional and requires an explicit scope-configuration selection or written authorization. It can cover default or weak credentials and previously leaked credentials appearing to relate to the customer, domains or users. Authorization can be limited by target, portal or credential category and withdrawn; cessation is promised within a commercially reasonable time. The customer needs the relevant rights and personal-data basis. The addendum also acknowledges possible account lockouts, alerts, password resets or similar effects.

Repository access and code retention

The AutoFix GitHub App requests read/write access to Code, PRs and Workflows, plus read access to Checks and Actions. This write-enabled remediation path differs from the default scanning integration's minimal read-only scope.

Aikido says repository analysis uses a fresh container and wipes analysis data when that container is terminated. It documents exceptions: AutoFix retains original and fixed diffs for the affected files, and AutoTriaged SAST findings retain call trees for up to two weeks.

Repository-data documentation identifies EU, US, US Gov, ME and AU application instances and describes storage in the corresponding regions.

AI processing, personal data and test records

Aikido describes AI operations as inference-only, with data processed transiently in memory rather than retained or reused, and says small anonymized code fragments may support accurate triage and fixing. For SAST, IaC and container AutoFix, it says only minimal required snippets are sent through encrypted channels to AWS Bedrock, and neither Aikido nor AWS Bedrock trains or fine-tunes on that code. These vendor statements concern the specified processing scopes.

The privacy policy places personal data in Europe and the United States and retains service data during the customer relationship and afterward for operations, history and archiving. Personal data under Aikido's control may be deleted following a verified request.

Pentest activities and findings are logged. Test data may support delivering results and improving the service; credential-testing credentials and other personal data are limited to delivering results under the DPA. Raw test data is retained for a limited period to answer follow-up questions or defend claims, then deleted; the cited clause gives no fixed duration. Aikido may retain an archival copy of the final report.

The MSA preserves customer ownership of content, repositories, codebases, containers and cloud infrastructure, while permitting processing needed to perform obligations, enforce contractual rights, or meet legal requirements or authorization. After termination, it allows 30 days to retrieve native-format content and usage analytics and calls for content to be purged within 200 days. Retention during that period is limited to backup availability.

FAQ

Q1. Is source-code access required for a pentest?

No. Aikido says providing repository access significantly improves context about application logic, roles and data flows, leading to deeper coverage and more accurate findings.

Q2. What must be connected before Continuous Pentesting?

An initial assessment must be completed and the repositories used to identify security-relevant changes must be connected. Once configured, Aikido can run targeted pentests as the application changes.

Q3. Who controls API and MCP access?

API token setup uses OAuth Client Credentials for server-to-server communication. A workspace admin first creates the client credentials used to obtain an access token.

MCP permissions are managed by workspace admins and apply to every user across MCP tools. Issue listing is limited to issues the user can already access, and ignoring findings remains subject to the user's existing role permissions.