FAQ
Short answers to what Studio users ask most, each with a link to the page that explains it; for a refusal, a failure card or a banner quoted word for word, use Troubleshooting.
Working in Studio
"You've reached the limit of 256 agents." (or knowledge graphs, data connectors, environments, workflows)
Each user can hold 256 of each type. Delete what you no longer use; every list row has Delete. See Limits and quotas.
"This PDF has N pages. Knowledge graphs accept PDFs of 32 pages or fewer."
A knowledge graph builds from one PDF of at most 32 pages and 50 MiB. Bring the PDF within both and choose it again.
My environment shows Needs attention
Read the reason on the environment ("Provisioning did not finish within 15 minutes" is one) and click Provision; it also retries after a failure. Until the environment is Active, agents wired to it cannot run code.
"An execution environment is required to activate."
A Specialist agent needs an environment on its Agent Map before Save & Activate. Supervisors work only inside Workflows and take none.
The agent's Create button stays disabled
Create Agent (v1) and Update agent enable once you choose Use as written, or Optimise Prompt then Use optimised or Keep mine; Revert unlocks it.
"This request needs a larger plan than a chat run can take on at once, so it did not start."
Narrow the request, split it across several chats and carry the results forward, or build it as a Workflow, which is made for larger multi-step work.
"Studio is updating" or "AlphaAgent Studio was updated. Refreshing to load the latest version…"
A release is being applied to the deployment. The first message (on knowledge-graph screens, ending "try again in a few minutes.") means a service is not yet back; wait and retry. The second reloads Studio on its own after four seconds; use Refresh now to reload at once.
I cannot delete a knowledge graph
"Agents pinned to any version keep the graph from being deleted." Move or unpin the agents first (Move N agents here on another Ready version, or change each agent's Agent Map), and cancel or wait for any building version.
I edited my agent but chat still behaves the old way
Editing creates a new agent version; the conversation keeps its active configuration until you activate the new version on the Agent Map. The ⓘ button in the composer shows what the conversation is running.
Do runs stop when I close the tab?
No. Runs execute on the server: leaving the page, closing the tab or losing your connection does not stop one, and Studio shows no prompt when you leave. The run continues in the background; come back to it from Recent Chats in the sidebar or from the Inbox, and the board is rebuilt from the run's record. Stop in the composer or on the run cancels one.
Why is Chat empty on a fresh Studio?
A fresh Studio has no agents. Create one under Agents before you start a conversation. See Getting started.
We already use Claude Desktop or Claude Code with skills and memories. Can we bring them?
Yes. A role skill becomes an agent, a procedure a workflow, references and team memories a knowledge graph, an MCP server a connector, and every one of them arrives versioned. Personal preferences stay with the person and local tooling moves into a cloud execution environment; the page explains each choice. See Bring your Claude skills and memories.
Limits
"That selection has more than 2000 files" or "zip expands to N MB; the limit is 500MB"
The message names the cap it hit: 100 MB per uploaded file, 2,000 entries and 500 MB expanded for a zip, 2,000 files and 500 MiB for a zip download. Split the selection or the zip and try again. A zip refusal is all-or-nothing: nothing from it lands. See Limits and quotas.
Assign is disabled with "No more users fit in this deployment (20 max)"
A deployment has 20 seats by default; a seat is a StudioUser grant, directly or through a group. Remove users who no longer need the deployment, or ask for a higher cap when the Organisation is configured. See Users, groups and roles.
The workflow API answers 429
Wait for the Retry-After header: seconds to the next minute for rate_limited, seconds to the next UTC day for quota_exceeded, 5 for too_many_active_runs. If the key's limits are too low for your integration, an administrator raises them with Edit on the key in Programmatic Access. See Limits, errors and codes.
A throttle strip is across the top of Studio
Amazon Bedrock in the deployment's AWS account is throttling; the strip names the model and the quota. Raise the Bedrock quota for that account and region with AWS. The three states and what happens to the run are on Troubleshooting.