Skip to main content

Limits and person-only steps

Synced from bicycle-studio-api

Source: studio_guide("limits") (GET /api/studio/v1/guide/limits) rendered at origin/platform (1afd83f), synced 2026-09-28. Do not edit this page here; change the source and run yarn sync:studio.

Steps only a person takes​

The tool answers with a link and changes nothing. Give the person the link, say what it changes, and wait.

StepWhoWhere
First publish of a function; a publish that widens it (capability, model, query, audience, MCP exposure, class, timeout, image, compute budget)a person with publish rights on functions, signed in/<company>/apps/functions/<name>?publish=vN
Publish a workflow revision with gated changes (a send, its approval, a new query, model or Use, AI text under auto, a higher compute budget)a person with publish rights on the workflow/<company>/apps/workflows/<wf>
Publish an app version that adds or re-pins an import (other versions publish at once with apps:publish)a person with publish rights on the appthe app's ?publish=vN link
Expose a function or workflow to agents or MCP; notify.on_success: sendthe owner, in Studioanswer person_required
Approve or reject a sendan approver named on the stepStudio approvals inbox
Mail policy, action policy ("may run on its own")a tenant adminStudio settings
Per-workspace agent capsBicycle staff

Sends​

  • At most 50 recipients a step; max_sends_per_run default 1, up to 500; 500 recipients per tenant per UTC day.
  • An approval waits at most 14 days. A key that was sent is never sent again.
  • On preview, mail goes only to the deployment allowlist (workflow_mail_policy says whether your address is on it); any other address is refused at save (recipient_not_allowed).
  • Action sends (Slack, PagerDuty, Jira, webhook) need a person's approval unless a tenant admin allowed that action to run on its own. Never test an action that pages or tickets real people.
  • A failed screenshot sends nothing.

Cost and time​

WhatLimit
llm workflow stepmax_cost_usd <= $25; a run estimated over $25 is refused (cost_cap)
llm.call from codemax_calls <= 50, max_usd <= $5
code timeout_ms<= 30 s sync; async up to the class's maximum run time (table below); over 900 s is a widening
workflow function stepthe function's own timeout; a snapshot 240 s
agentdefault 30 steps, 80 tool calls, $2, 600 s; ceiling 100, 300, $10, 1800 s
fn.call depth, agent nesting3 each

Sizes​

WhatLimit
app manifest32 queries, 16 imports, 16 blobs, 20 tabs
declared querySQL <= 8000 chars, maxLimit default 1000, at most 100000
app bundle25 MiB zip, 2 MiB per asset
function package, blobsthe code sandbox below
MCP-exposed function schema<= 8 KiB, <= 16 top-level properties, depth <= 4
function name^[a-z][a-z0-9_]{0,55}$
workflowworkflow.json <= 256 KiB, a file <= 1 MiB, a json artifact <= 1 MiB, 32 blobs, 32 targets

Data and trust​

  • An app call runs as the viewer, an MCP call as the token's person, a scheduled workflow as the tenant's service identity. Each reads only what that identity may read.
  • You see only your tenant's objects; others answer 404.
  • Agents only read. Writes happen only in workflow send steps.
  • Deleting is disabling (functions, workflows); there is no hard delete.
  • No customer model ids, metric names or data in the public template or docs; use <model>.

The code sandbox (generated from the runtime and Studio)​

The runtime did not answer just now, so the image's packages, classes and timeouts are not listed here: function_kinds shows them. There is no pip install and no network in the sandbox.

Caps Studio enforces:

  • An invocation's output is at most 6 MiB; at most 200 capability calls per invocation.
  • ctx.query(...) answers at most max_rows rows (default 1,000), and never more than the declared query's maxLimit or 10,000.
  • A workflow query node answers at most 10,000 rows, and at most the declared query's maxLimit (default 1,000, up to 100,000); rows past the cap are dropped without an error, so set maxLimit and aggregate in SQL (for example by day) when history is long.

Daily compute budget: compute units are wall seconds x the class's vCPU (a call reserves its timeout_ms x vCPU and is settled to what ran). By default a function may spend 2,000 a day and a workflow's function nodes together 5,000. limits: {"compute_units_per_day": N} in function.json or in the workflow document sets its own, at most 2,000 on this deployment unless your workspace has its own maximum (more is refused in words). Raising it above what the published version had widens it: the publish review shows the change and a person confirms it in Studio; an agent cannot. A raise counts once published: a Try, a draft's run and a workflow try run get at most the default. When a budget is used up, calls answer quota_exhausted until the next UTC day. A long run is charged like any other: a function with a long timeout_ms (up to its class's maximum run time) reserves timeout x vCPU per call, so its budget must cover at least one call (validation warns when it does not).

guide_version bb2461328de4