How we look after your company's data.
This page sets out what Honeycomb does today to keep your company's data safe, the documents you can read or ask for, and every third party that processes your data for us. Where something is not in place yet, it says so.
System status
Honeycomb is up, but some work is running late.
- AppWorking normally
- DatabaseWorking normally
- Agents and schedulesRunning late
Uptime history comes with our status monitor.
Checked 2026-11-12 03:23 UTC. Full status page
Certifications
We do not hold a security certification such as SOC 2 or ISO 27001 yet, so you will not see a badge here. SOC 2 Type II is on our roadmap and is not under way. Until then, here is what Honeycomb does today, built into the product.
What we do today.
The same practices as on our home page, built into the product.
Read-only by default
Agents read your systems and prepare the work from day one. When work needs to write, your company decides: high-risk actions such as deletes, refunds and sends start switched off, and changes go to a person for approval or follow approval rules your company sets.
Read-firstPer-company isolation
Every fact, run and approval belongs to your company, and our servers only return it to people signed in to your company. Uploaded files are encrypted with a key per company.
Company-scoped accessEncrypted connector credentials
Every connector token and secret is encrypted with AES-256-GCM before it is stored. Call records keep the facts of a call, never its data or credentials.
AES-256-GCMScoped app tokens
Apps from the Software Factory never hold your passwords. They get short-lived tokens for approved systems only, and their write requests wait for approval.
15-minute tokensEvery step on the record
Every AI call, approval and change to what the brain knows is traced: who asked, what happened, when and why. Whenever you look back, you see exactly what happened and why.
Full audit trailYour company holds the controls
Your company sets its own usage limits. The moment one is reached, work stops before the next AI call goes out, and any running agent can be stopped on the spot. Control stays with you.
Owner-set limitsInvite-only access and SSO
Sign-up is by invitation. Companies can sign people in through their own identity provider and turn off passwords for their domain.
OIDC single sign-onOutside data stays data
Data read from connected systems is clearly marked and placed after a trust preamble, so the AI treats it as data, never as instructions.
Injection mitigations
Documents
Read the policies here. Ask for the rest and we will send them to you.
- PublicRead the Privacy Policy
Privacy Policy
What data we hold, our roles, AI providers and retention.
- PublicRead the Terms of Service
Terms of Service
The terms for companies using Honeycomb, including how we process data for you.
- PublicRead the Cookie Policy
Cookie Policy
The sign-in cookies and the preferences kept in your browser.
- On requestRequest the Data processing terms
Data processing terms
Our data processing terms as a separate document, for your review.
- On requestRequest the Security overview
Security overview
How Honeycomb is built and run, for your security review.
How AI uses your data.
No training on your data. Our AI providers do not use the data we send them to train their models, and we do not use your data to train models.
If your company brings its own AI provider key, retention for those AI calls follows your own agreement with that provider. Otherwise, it follows each AI provider's standard API data retention terms, as described in our Privacy Policy.
Subprocessors
The third parties that process customer data for us on our hosted service. Last checked 2026-11-11.
Our subprocessors
- What it does
- Hosts Honeycomb: the web app, its server functions and scheduled jobs, and the server logs.
- Customer data
- Everything the app receives and sends while it runs, and server logs (the page or API path, and error lines with personal details removed).
- Where
- United States
- What it does
- The Postgres database where Honeycomb stores your company's data.
- Customer data
- Accounts and companies, the company brain, uploaded files (encrypted with a key per company), connector credentials (encrypted), agent runs, approvals and usage records.
- Where
- United States
- What it does
- The default AI model provider (Claude): answers, drafts, agent runs, building the company brain from what you connect and upload, and web search when it is on.
- Customer data
- The text each request needs: the question or instruction, the relevant company knowledge and the data read from connected systems. Web search queries when web search is on (a company can turn it off).
- Where
- See vendor
- What it does
- Search embeddings for the company brain (text-embedding-3-small), for every company; also answers and web search when a company chooses OpenAI as its AI provider.
- Customer data
- The text of company knowledge and of searches, turned into embeddings. When chosen for answers: the text each request needs: the question or instruction, the relevant company knowledge and the data read from connected systems.
- Where
- See vendor
- What it does
- Sends Honeycomb's account and security email: sign-up invitations, password resets, account-lock notices and sign-in alerts. We send no marketing email.
- Customer data
- The recipient's name and email address, and the message content.
- Where
- United States
Google Workspace
Privacy policy (Google Workspace, opens in a new tab)- What it does
- Our own company mailboxes, where the customer and privacy requests sent to our contact address arrive and where we reply.
- Customer data
- The emails people send us and our replies.
- Where
- United States
Optional AI providers
These are used only when a company chooses one in Settings, AI model, with its own API key. A company that brings its own key for any provider, Anthropic and OpenAI included, works under its own agreement with that provider.
Google (Gemini API)
Gemini API terms (Google (Gemini API), opens in a new tab)- What it does
- Answers, drafts and agent runs, when a company chooses Gemini with its own API key.
- Customer data
- The text each request needs: the question or instruction, the relevant company knowledge and the data read from connected systems.
- Where
- See vendor
- What it does
- Answers, drafts and agent runs, when a company chooses Mistral with its own API key.
- Customer data
- The text each request needs: the question or instruction, the relevant company knowledge and the data read from connected systems.
- Where
- See vendor
- What it does
- Answers, drafts and agent runs, when a company chooses DeepSeek with its own API key.
- Customer data
- The text each request needs: the question or instruction, the relevant company knowledge and the data read from connected systems.
- Where
- See vendor
- What it does
- Answers, drafts and agent runs, when a company chooses Groq with its own API key.
- Customer data
- The text each request needs: the question or instruction, the relevant company knowledge and the data read from connected systems.
- Where
- See vendor
Systems your company connects
Systems your company connects, such as QuickBooks or FactSet, and your own sign-in provider for single sign-on, are your company's own vendors, not our subprocessors: Honeycomb works with them on your behalf, with the access you grant. See integrations
Changes to this list
We give 30 days' notice before a new subprocessor starts processing your data, so your company can object. Get notified of changes
Contact and responsible disclosure
If you think you have found a security issue in Honeycomb, email hello@scaledabundance.com with what you found and how to reproduce it. Please do not access other companies' data or disrupt the service while you look.
For anything else about this page, write to hello@scaledabundance.com.