Skip to content
KitploitKITPLOIT
ToolsExploitsBlog
Log in
Submit
ToolsExploitsBlog
Submit

Hacking, PenTest, and Cybersecurity Tools for Your Security Arsenal!

Kitploit is a directory of hacking, cybersecurity, and pentesting tools. Discover the latest project updates to find vulnerabilities, analyze systems, automate testing, and strengthen your security.

FeedsContactPrivacy© 2026 Kitploit

Tool Directory

Categories

View all categories
Loading categories
chatwoot-agentbot-global-token — Chatwoot v4.19.1 - High: 8.7 - Remote admin - token disclosure / cross-tenant object access - inbox agent-bot view returns a global AgentBot access token | Kitploit
Tools/GitHubGitHub/abraxas/chatwoot-agentbot-global-token
Vulnerability AnalysisExploitationWeb Application ExploitationInformation GatheringWeb SecurityPenetration TestingLabs & Practice
GitHub
abraxas/chatwoot-agentbot-global-token

chatwoot-agentbot-global-token

Chatwoot v4.19.1 - High: 8.7 - Remote admin - token disclosure / cross-tenant object access - inbox agent-bot view returns a global AgentBot access token

View Repository
15 days agoNot yet reviewed
Share

Abraxas Labs - chatwoot-agentbot-global-token

chatwoot-agentbot-global-token

A Chatwoot administrator can attach an instance-wide AgentBot to their own inbox, then read that bot's API token from the inbox view. The token is meant to stay hidden because the bot belongs to no tenant. With it, the attacker reads and writes conversations in every other tenant that uses the same bot.

Class: Token disclosure / cross-tenant object access Reach: Remote, authenticated account administrator

Chatwoot v4.19.1 - caab07a8

IDno CVE yet
ClassToken disclosure / cross-tenant object access
ReachRemote (administrator of any account)
CWECWE-200 / CWE-863
CVSSHigh: 8.7 CVSS:3.1/AV:N/AC:L/PR:H/UI:N/S:C/C:H/I:H/A:N
ProductChatwoot
Affectedv4.19.1 caab07a8f00032d5205b5a5d7e2eb0ba2f3d2d82
Authadministrator of any account on the instance
LicenseGNU Affero GPL v3.0
Labloopback Docker compose; Postgres, Redis, Puma, Sidekiq

The score assumes multi-tenant hosting, which is Chatwoot's own deployment model: one instance, many customer accounts, each with its own administrators. On a single-tenant install there is no boundary for the token to cross, and the same request only hands an operator a credential they could already read from the console.

What an attacker can do

Starts as an administrator of any account. GET /api/v1/accounts/:id/agent_bots already lists global bots, but the presenter blanks their token. The inbox view does not use the presenter.

PrivilegeMethod / sinkRecon / integrity / availabilityLab-proven or inferred
See the global bot in the tenant listGET /api/v1/accounts/:id/agent_botsReconLab-proven. listed_token=None
Attach the global bot to an inbox the attacker controlsPOST /api/v1/accounts/:id/inboxes/:inbox/set_agent_botIntegrity (config change)Lab-proven. InboxPolicy#set_agent_bot? is administrator-only; AgentBot.accessible_to admits account_id IS NULL
Read the bot's access tokenGET /api/v1/accounts/:id/inboxes/:inbox/agent_botRecon (credential)Lab-proven. leaked_token_len=24
Read a victim tenant's conversationGET /api/v1/accounts/:victim/conversations/:id with api_access_tokenRecon (message content)Lab-proven. Witness returned
Post into a victim tenant's conversationPOST /api/v1/accounts/:victim/conversations/:id/messagesIntegrity (customer-visible message)Lab-proven. Victim session reads it back
Change status, priority, assignee, labelsPATCH conversation, POST /conversations/:id/labelsIntegrityInferred from BOT_ACCESSIBLE_ENDPOINTS; same token and policy
Enumerate other tenants served by the botagent_bot_inboxes rowsReconInferred. The bot is admitted to every account where it has an inbox
Negative control: attacker's own user token on the victim conversationGET victim conversation with the user's access-tokenn/aLab-proven. 401
Negative control: token in the list viewGET /agent_botsn/aLab-proven. access_token absent
Negative control: non-administrator account membersame inbox viewn/aSource-proven. The partial gates the field on Current.account_user&.administrator?

Stays closed on this leftover

AxisMethod / sinkWhy it is closed
ReconVictim inbox read by the attacker's user tokenInboxPolicy#show? is account-scoped; lab returns 401
ReconToken on every other agent-bot viewAgentBotPresenter blanks it when account_id is blank
ReconToken for a non-administrator memberCurrent.account_user&.administrator? in the partial
IntegrityAttaching a bot to another tenant's inboxset_agent_bot writes only to Current.account inboxes
IntegrityCreating a global bot from the dashboardNo UI or API path; bots are created inside an account

How I found it

I pinned v4.19.1 (caab07a8) and read the inbox controller against the agent-bot controller. Every agent-bot view wraps the record in AgentBotPresenter. The inbox view hands the raw record to the partial:

# app/controllers/api/v1/accounts/inboxes_controller.rb
def agent_bot
  @agent_bot = @inbox.agent_bot
end

The presenter exists to hide one thing:

# app/presenters/agent_bot_presenter.rb
def access_token
  return if account_id.blank?
  Current.account.id == account_id ? super&.token : nil
end

account_id.blank? is exactly the global-bot case. The partial still gates on administrator, so a regular agent gets the record without the token. An administrator gets the token.

Wrong turns: I first assumed the bot list leaked it too, which it does not; and I assumed the attacker needed a bot of their own, when the global bot is the one that crosses tenants. The lab keeps both as controls.

Lab

python3 chatwoot-agentbot-global-token-Abraxas-Labs.py

or:

cd lab
./run.sh

Brings up Postgres 16 with pgvector, Redis 7, Puma and Sidekiq from chatwoot/chatwoot:v4.19.1, seeds two tenants and a global bot, then drives the API over loopback on 127.0.0.1:28210. run.sh tears the stack down on exit.

IOC listed_bots=1 global_bot_id=1 listed_token=None
IOC attached global_bot=1 to inbox=1
IOC leaked_token_prefix=FV94xdoC1mAL token_len=24
IOC control attacker_user_reads_victim status=401
IOC bot_token_read status=200 account=2 conversation=1
IOC bot_token_write status=200 victim_confirms=true
SUCCESS CHATWOOT-F1-AGENTBOT-TOKEN account=1 inbox=1 global_bot=1 leaked_token_len=24 read_victim_account=2 conversation=1 VICTIM-SECRET-CONVERSATION-BODY

The token is 24 characters because Chatwoot's has_secure_token default length is 24. The value changes per seed; only the shape is fixed.

The fix

Pass the presenter in app/views/api/v1/accounts/inboxes/agent_bot.json.jbuilder:

json.agent_bot do
  json.partial! 'api/v1/models/agent_bot', formats: [:json],
                resource: AgentBotPresenter.new(@agent_bot) if @agent_bot.present?
end

A second layer is worth having: refuse set_agent_bot for bots whose account_id is not the current account, unless the instance deliberately shares that bot. The presenter is the narrow fix; the scope is the durable one.

References

  • chatwoot/chatwoot caab07a8f00032d5205b5a5d7e2eb0ba2f3d2d82
  • inboxes_controller.rb
  • agent_bot_presenter.rb
  • _agent_bot.json.jbuilder
  • CWE-200 · CWE-863

License

GNU Affero GPL v3.0. See LICENSE. Loopback lab only. No warranty.

Abraxas Labs [email protected]

Download Tool

Most Popular

View all →

Discover the most used tools by our community.

  • 1

    OsintHunter

    letox4/osinthunter

    🎯 The ultimate OSINT toolkit: investigation graph, confidence scoring, alias families, map, timeline and reports. 49 modules, 400+ sites, no API keys required.

    831
  • 2

    ghostlock-app

    yukonga/ghostlock-app

    GhostLock One-Tap Execution App (CVE-2026-43499)

    534
  • 3

    ARTEX

    baarai/artex

    LLM multi-agent autonomous penetration testing platform with planner/worker agents, asset and exploration graphs, MITM traffic recording, and vulnerability retesting.

    429
  • 4

    redforge

    cybersecurity91/redforge

    Browser-side red team toolkit with 102 offline tools for recon, payload generation, credential attacks, crypto, encoding and engagement reporting.

    355
  • 5

    looking-glass

    anonghosty/looking-glass

    CSIRT platform that ingests Shadowserver reports into MongoDB and replays them as interactive maps, analysis, findings, advisories and executive dashboards.

    284
  • 6

    CyberStrikeAI

    ed1s0nz/cyberstrikeai

    The system of action for AI-native cybersecurity—where intent becomes governed execution, evidence becomes operational memory, and every operation improves the next.

    275

Explore all tools

Browse our collection of tools

View all tools →