Migrate auth to magic link
P1Ship Squad
everywhere · BYO compute
Run agents on your own machine.
A tiny daemon pairs Stackon to your laptop, GPU box, or self-hosted runner over a one-time code. Jobs execute locally with your own model keys — only the trace metadata flows back. No model traffic leaves your network.
bridge · pairedzero egress
stackon
orchestration
queues job →
ABCD-EFGHjob →← trace
your box
jobs run here
your keys · local
trace · status ok · bridge_id set
└─ agent.run
└─ llm.completion
runner: bridge-v1online · 30s ♥
Pair by 8-char codeYour laptop, GPU, or VPCTraces still flow back
01
Pair in one line, no install ritual
Generate a pairing code in Dashboard → Compute and run `npx @stackon/bridge pair --code …` on the target machine. The one-time, 15-minute code trades for a long-lived bridge token stored at chmod 600 — share the code like a join link, treat the token like an API key. No package to vendor, no agent to babysit.
bridge · pairedzero egress
stackon
orchestration
queues job →
ABCD-EFGHjob →← trace
your box
jobs run here
your keys · local
trace · status ok · bridge_id set
└─ agent.run
└─ llm.completion
runner: bridge-v1online · 30s ♥
02
Your compute, your keys, your data
The daemon calls the Anthropic API directly with the ANTHROPIC_API_KEY set on the host — it never receives your team's stored keys. Prompts, code, and model responses stay on hardware you control. Stackon only ever sees the trace it gets back, so sensitive work never crosses your perimeter.
compliance · trust layeraudit-ready
agent.coderAgent run· trace · 8c4f21a12:04:11
u · danaRole changed· member · owner12:04:42
proxyPII redacted· 3 replacements12:05:09
u · renBYOK key rotated· anthropic12:06:30
pii proxy · standardscrubbing
in email dana@acme.io, card 4242 4242 4242 4242
out email <REDACTED:email>, card <REDACTED:credit_card>
anthropic…aF3kopenai…9Qx2
AES-256-GCM03
Runs land in Trace like any other
Each completed job inserts a trace with an agent.run root span and a nested llm.completion span — tagged runner: "bridge-v1" so you can tell local runs from cloud runs at a glance. Token counts, per-model cost, and duration all roll up exactly as they do for cloud runs, replayable in the same viewer.
trace · run_8c4fok · 742ms · $0.0053
agent.plan742ms
tools.search_code86ms
llm.complete_refactor612ms
tools.edit_file78ms
evals.no_regression54ms
agentllmtooleval5 spans · 3,007 tok
04
Dispatch from the runner, watch it locally
Once a bridge is online — heartbeating every 30s, polling for jobs every 3s — it shows up in the Agent runner's "Run on" dropdown. Pick it, dispatch a run, and the work executes on your box while the live job and its status surface back in the Compute dashboard. Revoke any host with one click to cut it off server-side.
missions · board4 lanes
Backlog2
Draft changelog entry
In progress1
Refactor billing webhook
Ship Squadrunning
Done1
Add OpenAPI spec
Docs CrewTrace
Archived1
Spike: vector search
@all ship the billing fixSend
npx, one line
Install
every 30s
Heartbeat
trace metadata only
Egress
Part of one platform
Bridge works hand in hand with Apps & Surfaces.
Speed plus trust — prove your agents got better this week.
Bridge is one piece of Stackon, the observability-first workspace for teams running Claude and Codex. Start free and instrument your first run today.