How BRIDGIT works
Your agent connects intent to hardware. You keep authority over the phone.
Today: a device manager
- Use your own Android 14+ phone with Tailscale enabled.
- Open BRIDGIT and pair a named agent with a separate key.
- Grant only the permissions and capabilities you want to use.
- Your agent can request snapshots, short audio clips and accessible gallery photos over the private network.
The current switches apply across all paired agents. The phone logs activity and lets you rotate or remove agent keys. Open BRIDGIT after restart or interruption to restore camera/microphone access; snapshots currently require an unlocked phone.
Next: agent-created appliances
- Tell a compatible agent what job you want the phone to do.
- The agent proposes a bounded workflow from supported device capabilities.
- Review the permissions, destinations and behavior on the phone.
- Activate the workflow. The proposed local runtime executes approved actions without requiring the agent to remain online.
Local execution does not make remote messages work without internet. Android restrictions, a powered-off phone or a lost connection can interrupt availability.
Designed for visible control
New grants require owner approval. Appliance status stays visible. Stop and revoke controls take priority over agent requests. These are v3 design requirements, not a claim that the current personal build implements every proposed control.
Where we are going
Camera appliances and a family help-station pilot are planned after the local workflow foundation. Baby monitoring requires dedicated streaming and connection-health work. Trigger-word help requires tested local detection, consented contacts and explicit failure handling.