> ## Documentation Index > Fetch the complete documentation index at: https://docs.cantina.security/llms.txt > Use this file to discover all available pages before exploring further. # Changelog > Product updates across Cantina, Apex and Clarion Product updates across Cantina Code, bounties, Apex and Clarion, from August 2024 to today. Related improvements are grouped, and features in beta or early access are labelled. **November 2026** ### Clarion * **Argus.** See the automated attacks reaching your apps through Vercel and Cloudflare, who is behind them and what Clarion caught, and query request logs directly. * **Vulnerability SLAs.** Set remediation SLAs and track due dates on issues. Dependabot alerts are grouped by package, and withdrawn advisories close on their own. * **Bounty triage (early access).** An AI triage agent reviews findings, forwards credible ones, opens fix pull requests, files tickets and flags duplicates. You can choose whether triage is handled by Cantina. * **Security hardening.** Connected MCP server credentials stay isolated, GitHub tokens no longer enter agent environments and GitHub connections are limited to what you can access. **3 November 2026** ### What We Shipped: More Context for Every Investigation Amazon Inspector findings link to assets and issues in Clarion, and agents can query live Inspector data with read-only tools. Apex Black assessments include a downloadable PDF report, and replies in Slack threads continue an issue's triage. [Read the post](https://www.cantina.security/blog/what-we-shipped-investigation-context) **October 2026** ### Clarion * **Connected evidence.** Start an investigation from any asset. The Brain links CVEs, advisories, package versions and affected resources, and an issue reopens when new evidence contradicts its resolution. * **Amazon Inspector.** Findings connect to assets and issues, are grouped by the fix they share and can be queried live with optional read-only access. * **Working in Slack.** Clarion follows the threads it's part of. Reply in an escalation or scheduled-run thread to continue the investigation, and attach PDFs, Word documents and text files in chat or Slack. * **Approvals and privacy.** Approval requests explain why the agent needs each tool and offer Always approve. Your data is pseudonymized before it reaches less trusted systems. * **Microsoft 365.** Approval-based remediation for compromised accounts, audits of MFA, OAuth consent and inbox rules, and search across SharePoint and OneDrive. * **Integrations.** Connect integrations from chat. New sources include CrowdStrike cloud posture risks, GitHub organization audit logs, Supabase auth audit logs, Wiz risk grouping and software bills of materials, with custom authentication headers for MCP servers. * **Analytics and filters.** Dashboards for pending approvals and questions and for escalated and open issues, detection templates and personal saved issue filters. * **Early access.** Agent Builder for creating agents in conversation, with shareable sessions, an organization console with issues across workspaces, and Clarion Bounties with identity verification, reward payouts, program terms, disputes and migration from Cantina. ### Apex * **Apex Flash-1.** A downloadable open-weights model for focused security investigations. * **Scoring and filtering.** A workspace CVSS 3.1 scoring policy with feedback, severity filtering and filtered Markdown exports. * **Coding agents.** A panel for connecting coding agents to Apex. * **Credits.** Automatic credit top-up, with consent requested again when prices change. * **Scan reliability.** Pull request and fix reviews use isolated, commit-pinned context. A new scan can't start while one is running, and Apex links you to the active scan. * **Apex Black.** Download assessment reports as PDFs through secure, short-lived links. ### Cantina * **Bounties on Clarion.** Bounty programs can move to Clarion, and migrated programs stay viewable in Cantina as read-only. **16 October 2026** ### What We Shipped: Introducing Apex Flash-1 Apex Flash-1, plus a conversational way to create Clarion agents, a workspace CVSS 3.1 scoring policy for prioritizing Apex findings, and investigations that start from an asset with context from the Brain. New Brain connections include Swif, Pylon, Supabase and Devin. [Read the post](https://www.cantina.security/blog/what-we-shipped-apex-flash-1) **1 October 2026** ### Introducing Apex Flash-1 Cantina and Yeta Labs released Apex Flash-1, an open-weight cybersecurity model built on GLM-5.3-Flash and post-trained for security work. The post covers training, evaluation results and where to download the model. [Read the post](https://www.cantina.security/blog/introducing-apex-flash) **September 2026** ### Clarion * **The Brain for everyone.** The Brain is available in every workspace. People from Google Workspace, Okta, Vercel, GitHub, Kandji and Vanta are linked to the assets and accounts they own. * **Integrations.** Amazon Inspector (early access), NinjaOne, Supabase, Pylon, Devin and Swif, plus CrowdStrike Next-Gen SIEM search, Vercel audit logs and two-way Wiz status sync. * **Issues.** A new Open status for non-urgent work, issue types, linked and duplicate issues, sub-issues and fewer repeat alerts. * **Collaboration.** Ask a teammate a question privately, give reasons when approving or rejecting, manage monitors and rules through MCP, authenticate custom webhooks and get richer Slack messages. Paused monitors are clearly marked. * **Detection and reporting.** Detection of credential harvesting in Claude Code sessions, and remediation SLA reports as PDFs. * **Workspace controls.** Choose the model each agent uses, create agents, skills and monitors from chat, dictate by voice and pause a whole workspace when needed. ### Apex * **Setup that learns.** Audit Setup and fix instructions reuse selected earlier review feedback, and workspaces and threat models can be set up in conversation. * **Fix reviews and findings.** Inline fix-review status, history and retries, better pull request search, bulk finding actions, live findings and Scan Insights. * **Budget controls.** Finer control over budgets and credits. ### Cantina * **Bounty submissions.** Every submission now includes a Risk section. * **Insights.** Bounty insights break results down by asset group, with true and false positive rates. * **Asset scope.** Assets can be hidden instead of deleted. * **Privacy.** Client and staff identities are anonymized in bounty notifications and exports. **29 September 2026** ### The Brain: giving security agents knowledge of your environment The Brain launches as the shared context layer for Cantina's security agents, linking identities, cloud assets, devices, repositories and vulnerabilities. The post walks through demo investigations and an observational comparison across 40 alerts. [Read the post](https://www.cantina.security/blog/the-brain-security-agents) **18 September 2026** ### What We Shipped: Claude Code Credential-Harvesting Detection Clarion flags Claude Code sessions that read a secret, such as a `.env` file, and then make an outbound request, as well as one-line exfiltration commands. The release also adds new integrations, remediation SLA reports as PDFs and Apex finding-review updates. [Read the post](https://www.cantina.security/blog/claude-code-credential-harvesting-detection) **August 2026** ### Clarion * **AI coding agent governance.** See Claude Code activity across your team, keep an inventory of the MCP servers in use, approve or deny them and get alerted to activity from accounts outside your company. Decisions made in Slack and Teams are recorded, and Slack destination permissions are enforced. * **Exposure checks.** Replit and Vercel exposure checks, dangling-subdomain detection and a package inventory. * **Integrations.** Vanta with a vendor risk reviewer, Wiz (now generally available), Microsoft Defender XDR, NightVision, Envoy and Replit. Devices from GCP, CrowdStrike, Kandji and Huntress sync into the Brain. * **Agents.** A community library of shared agents and skills, handoffs to coding agents, an Exposure agent, weekly Slack triage summaries and multiple notification destinations. * **Issue closure and audit.** Closing an issue requires a verdict and a comment, issues can be closed from Slack and issue views are recorded in the audit log. ### Apex * **Ticketing.** Linear and Jira connections, finding export and reopening, and routing per workspace. * **Audit Setup (beta).** Review generated context and scope before a scan is approved. * **Scan accuracy.** CLI and pull request scans are pinned to exact commits, pull request and Fix Review runs no longer wait for workspace scans, and findings you marked invalid aren't reported again. * **Visibility.** Scan time and cost, service API credit balance, related-finding links and auditor feedback synced into client findings. * **Apex Lite.** Wider early access and isolated collaboration workspaces. ### Cantina * **Apex scans.** Start Apex scans on Cantina repositories and read the results through the API. * **Reward decisions.** Bounty reward decisions are explicit, including rewards marked "to be disclosed". **26 August 2026** ### Claude Code and MCP Telemetry in Clarion Clarion adds visibility into Claude Code and connected MCP servers, human approvals in Slack and Teams, and external exposure sweeps. Apex adds Audit Setup (beta), CLI updates and scan controls. [Read the post](https://www.cantina.security/blog/claude-code-mcp-security) **July 2026** ### Clarion * **The Brain (beta).** Clarion maps your assets across providers, learns your business context with your confirmation and builds a knowledge base from Jira, Linear and Notion that agents use during investigations. * **Roles and audit.** A Member role for day-to-day operational work, and the audit log shows before and after values for every change. * **Bring your own models (beta).** Use your own AI provider, with automatic failover between models. * **Integrations.** Notion, Huntress, HaloPSA, incident.io and Wiz. Okta and Entra remediation runs with approval, CrowdStrike detections become Issues and AWS onboarding uses CloudFormation. MSPs can private-label the Teams app. * **Built-in agents.** Vendor security questionnaires, SOC 2 evidence collection and Azure security posture review, plus a risk assessment agent in early access. * **Issues.** Bulk actions, PDF export and testing detection rules against past events before turning them on. ### Apex * **Apex Lite (beta).** Dedicated credits and personal workspaces for researchers. * **Service API.** Create workspaces, start and cancel scans and manage service keys through the API. * **Automation.** Scheduled Fix Reviews and scans, CLI scans of local source code, workspace-wide finding retrieval, and exports that include review feedback and comments, with proof-of-concept source optional. * **Workspaces.** Archive and restore workspaces and see Scan Insights. Pull request scanning is available to all companies. ### Cantina * **Cantina Managed.** Findings and comments from Cantina's internal researchers are shown as Cantina Managed. * **Bounty privacy.** Cantina staff names are hidden from researchers, and access to legal identity details is restricted. * **Account security.** Changing your email address requires signing in again. * **Bounty triage.** Bounties show whether they're triaged by Cantina or self-triaged. **June 2026** ### Clarion * **Clarion MCP server.** Connect your own AI tools to Clarion with workspace API tokens or OAuth, and manage issues, monitors, agents, skills and schedules from them. * **Audit log and roles.** Every action in the app, Slack, Teams and MCP is recorded, including the tools agents use. Admin and Read-only roles control who can change what. * **Scheduled tasks.** Run agents on a schedule. * **Issues.** Alerts and incidents now come together as Issues. * **Compliance reports.** Generate compliance reports from workspace analytics. * **Integrations.** Splunk, RSS and status-page monitors, Datadog Case Management, Tailscale tools and Cloudflare DDoS investigation. AWS now supports multiple regions and organizations, and pull requests are tracked on the issues they fix. ### Apex * **Fix Review generally available.** Available to company workspaces, with overlapping runs and credits counted per pull request. GitLab merge requests and current-repository reviews are supported, and fixes show as pending until they're merged. * **Findings and access.** Saved finding filters, timeline grouping, device login for the CLI and MCP, scan scheduling, scan permissions and fix instructions in reports. ### Cantina * **Audit log for everyone.** The audit log is turned on for every company. * **Access controls.** API key management is limited to admins and clients, and you're asked to confirm before private repository details are shown. * **Bounty submissions.** Researchers have submission limits, and findings can be filtered by asset. **May 2026** ### Clarion * **Workspace approval.** New workspaces are reviewed before they become active, and agent environments never hold AI provider keys. * **Response workflow.** On-call rotations, approve and reject buttons in Slack and Teams, verified phone numbers for alerts, PagerDuty notifications and chats started by mentioning Clarion in Slack. * **Integrations.** Linear, Kandji, Tailscale, SentinelOne, Microsoft Sentinel, GCP Cloud Monitoring and Datadog, plus read-only AWS investigation tools across multiple accounts. Agents can also work with custom APIs and CLIs you describe. ### Apex * **Audit scans and threat models.** Audit scans are available to every entitled customer, alongside reusable, editable threat models. * **More flexible fix reviews.** Review fixes against the current repository without a pull request, review selected findings, override branches and post results as GitHub comments. PDF exports run in the background. * **Security hardening.** Stronger authorization for webhooks and callbacks, narrowly scoped GitHub tokens, workspace isolation, private shared reports and protection against credit misuse. * **Human review.** Reviewer feedback and severity updates are recorded on findings. * **Apex Black (early access).** Assess several targets at once, with or without authentication. ### Cantina * **Card payments.** Pay with a card through Stripe. * **Entra SSO connections.** Microsoft Entra can be added as an SSO connection. **April 2026** ### Clarion * **Integrations.** Vercel, Jira, CrowdStrike Falcon with response actions, Microsoft Teams as an agent tool, Entra SSO and generic webhooks. * **Safer agent actions.** Agent commands are checked against permission tiers and reviewed for safety before they run. Failed investigations retry automatically. * **Chat and setup.** Continue a conversation with an agent after its investigation, share chats with teammates, set workspace-wide instructions and review connected GitHub accounts for security issues. ### Apex * **Fix Review.** Validate fixes, track pull requests and retest them, including fixes spread across several pull requests. * **Proof-of-concept replays and reports.** Watch proof-of-concept terminal replays and export PDF reports. * **Setup and CLI.** Guided onboarding, threat-model-guided setup, automatic GitLab scanning and a CLI that works without installing the GitHub app. Findings can be handed to Clarion in bulk. * **Early access.** Pull request scanning and Apex Black. ### Cantina * **Clarion connection.** New bounty findings can flow into Clarion. * **Single sign-on.** Microsoft Entra (Azure AD) SSO support. * **API keys and disputes.** Choose scopes when creating API keys, and track bounty disputes through more detailed states. **March 2026** ### Clarion * **Signals and detection rules.** Clarion correlates events across sources and raises alerts from detection rules, which AI can help you write. New detections cover Okta brute-force attempts, supply-chain attacks and expiring TLS certificates. * **Integrations.** AWS (CloudTrail, CloudWatch, GuardDuty), Google Workspace including Gmail investigation tools, GCP, Grafana, Better Stack, JumpCloud, 1Password, custom MCP servers, Cantina bug bounties and Apex. * **Skills and playbook templates.** Reusable agent skills and ready-made playbooks for account takeover, data exfiltration, email threats and security setting changes. * **Team workflows.** Incident assignment with email and in-app notifications, workspace invitations and public documentation. ### Apex * **Ultra scans.** Deeper Ultra scans became available for customer evaluation, with stronger protection against dependency attacks. ### Cantina * **Repository roles.** Client users can no longer be given reviewer roles. **February 2026** ### Clarion * **Clarion launches.** Sign in with your Cantina account, create workspaces, invite your team and track Alerts and Incidents with severity, history and follow-up tasks. * **AI triage agents.** Agents investigate alerts using your playbooks, ask for approval before taking sensitive actions and can be generated for you during workspace setup. A built-in assistant chat and an incident timeline are also available. * **Integrations.** GitHub, Slack, Microsoft Teams, Telegram, phone and SMS alerts, Okta, Falco, Cloudflare, Hypernative, Hexagate and Tenderly. DNS monitoring checks SPF, DMARC and DKIM records. ### Cantina * **Payments and access.** Stake balances are now handled through Stripe, and API keys support read and write scopes. * **Bounty collaboration.** New bounty notifications, unread reply badges on findings and client names hidden in bounty comments. * **Privacy.** Public pages now include a consent manager, and analytics only run after visitors give consent. **January 2026** ### Apex * **Repository scanning.** Connect GitHub to get started, scan private repositories and choose what's in scope. Repositories are categorized automatically and Git submodules are included. * **Workspaces.** Readable workspace URLs, consistent workspace editing and client comments on findings. **October 2025** Incident handling and controls Added Safe Harbor activation during bounty setup, notes on findings, a redesigned client dashboard, explicit repository membership instead of automatically adding all company members, incident escalation from critical bounty findings, personal and shared saved views, 48-hour reminders and optional refundable submission deposits. **September 2025** Triage at scale Added advanced filtering, centralized bounty triage, faster AI duplicate resolution, unread-comment tracking, simpler bounty confirmation, rejection and payment workflows, expanded role-based label controls, instant and daily notification options and clearer emails. **4 August 2025** Access and traceability Added individual bounty membership management, a centralized account and repository audit log, scoped-asset selection during submission and updated public link previews. **15 June 2025** Account security Added active-session visibility and individual session extension and termination, plus priority-based default bounty filters. **30 May 2025** Scope and discovery Added structured and nested bounty assets, explicit exclusions and severity-based rewards; expanded insights filters; enabled comment-to-finding conversion; launched the unified Opportunities page. **15 May 2025** Collaboration and automation Expanded mentions and default client notifications, unified engagement dashboards, and added Zapier triggers for label changes. **1 May 2025** Bounty self-service Added visible finding rewards, redesigned bounty pages, user email changes, cross-company repository invitations and role management, editable bounty settings with an audit trail, and client confirmation, rejection and labeling controls. **Mid-April 2025** Profiles and workflow Expanded researcher profiles, improved large-competition filtering, enriched Zapier finding data and sorted payments newest first. **1 April 2025** Review and reporting Added web audit reports alongside PDFs, redesigned code review with inline comments and findings, and exposed competition submission counts. **15 March 2025** Triage and integrations Added hidden researcher comments for judges and sponsors, AI-assisted finding checks, bounty insights, multiple company managers and OpsGenie alerts. **1 March 2025** Assignment Added finding assignment for customers, repository owners and judges; changed shortcut modifiers to prevent accidental status updates; accelerated findings lists. **15 February 2025** Onboarding and speed Added company-managed email-domain autojoin and improved platform performance. **1 February 2025** Visibility and access Made competition findings visible during judging where enabled; improved duplicate grouping; extended login sessions; listed private competitions with restricted or redacted information; added quicker repository access. **January 2025** Finding organization Redesigned the findings list, added emoji reactions and custom labels in private reviews, and introduced AI-suggested duplicates. **Late December 2024** Notification controls Added global and per-review notification preferences, with clearer descriptions of finding changes. **16 December 2024** Client prioritization Added a recommended-findings view using status, judge activity, escalations and researcher reputation. **3 December 2024** Navigation and privacy Added code-navigation shortcuts and private bounty-triager comments; changed ping eligibility and reduced default notification emails. **18 November 2024** Collaboration Added comment editing and resolution, improved submission templates, and finding and duplicate lookup by number. **23 October 2024** Review permissions Added client bounty triaging, full finding-status control for collaborative reviewers, team label management and quicker quality-field updates. **Early October 2024** Review experience Introduced Cantina Assistant to beta testers, tested a dedicated finding-creation page, and improved Mermaid loading, Markdown reliability and navigation. **2 September 2024** Submissions and notifications Added repository-specific project-ping controls, structured escalation and finding templates, "Select All" status filtering, and faster notifications. **26 August 2024** Content and performance Enabled image uploads in Markdown findings and comments, clarified payment disputes, added profile link previews and website resource links, and improved query performance. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.