Help
Troubleshooting, FAQ, and how to reach a human.
Troubleshooting
“Authentication required” on every page
Your session cookie expired or the server lost the row. Sign in again. If it happens repeatedly, your browser is blocking third-party cookies for our domain. Allow them or switch browsers.
Agent stuck in “Starting”
Container provisioning failed. Open the agent and click Resume, it’ll retry. If it fails twice in a row, check that you have an LLM key configured under Dashboard → LLM.
API key decryption failed
The account secret changed since you stored the key. Delete the affected key from Dashboard → LLM and add it again.
Connection keeps disconnecting
The OAuth refresh failed, usually because the provider revoked the grant or rotated their app. Disconnect, then reconnect. If it keeps happening, check the provider’s security settings for “recent logins.”
Task stopped running
Three consecutive failures auto-pauses a task. Open Tasks, find the red row, read the failure reason, fix the cause, and resume.
iMessage / SMS replies aren’t coming through
Confirm the trigger is still enabled in Settings (sometimes Apple flags a number as spam and we mark it inactive). If the trigger is on, send the agent “ping” over the web and confirm it’s running. If web works but iMessage doesn’t, the carrier hop is at fault; email us.
Web browser tool keeps getting blocked
Most likely the site fingerprinted the browser session as non-human. Set up a persistent browser session with a real login; the stealth layer is most effective when combined with real session cookies.
FAQ
- Can I run agents locally?
- No. Agents always run on our infrastructure. The tradeoff is zero-setup; the cost is you can’t air-gap.
- Do you train on my data?
- No. Your conversations, memory, and tool I/O are not used for any model training. We pass them through to the LLM provider you configured and store them encrypted at rest.
- What model is the default?
- Claude Sonnet (latest). Per-agent and per-message overrides are supported.
- Can I export my agent?
- Yes. Each agent has an Export button in Settings that bundles system prompt, memory, skills, tools, and tasks into a JSON archive. Re-import on another account to clone the agent there.
- Can I self-host?
- Not on our roadmap for early access. Talk to us if this is a hard requirement, schedule a call.
- How do I share an agent with my team?
- Team workspaces are coming. For early access, export the agent JSON and import it on a teammate’s account; both will run independently with the same configuration.
- What happens if my LLM provider has an outage?
- The agent’s task runs and chat replies fail with a clear error. If you have multiple providers configured, set a fallback in the agent’s Settings and we’ll route there automatically.
- Can the agent send physical mail / make payments?
- Not natively. Wire those up via a custom tool that calls your payments or mail provider; the agent can invoke it like any other tool.
Contact
Backbend is early access; the founder reads every message.
- Email: bradshaw.hyrum@gmail.com
- Schedule a call: cal.com/hyrum-bradshaw/backbend
- X / Twitter: @hyrumjb3