Letta API Outage
Timeline · 2 updates
- investigating Aug 17, 2026, 07:35 PM UTC
Investigating continued issues with Letta API
- resolved Aug 17, 2026, 08:39 PM UTC
resolved
Letta had 18 outages in the last 2 years totaling 8h 41m of downtime — averaging 0.7 incidents per month.
There were 18 Letta outages since January 22, 2026 totaling 8h 41m of downtime. Each is summarised below — incident details, duration, and resolution information.
Investigating continued issues with Letta API
resolved
We are investigating an issue related to errors relating to: cannot execute UPDATE in a read-only transaction
We have resolved!
app.letta.com recovered
General API went down
Some users are experiencing issues with sending messages with Letta API
Resolved
Issues with messaging API
Messaging API currently unavailable
Resolved
We have reports that users can not send requests on the Letta platform
Inference API went down
Degraded service on inference for non-byok users
Users might experience seeing 404 agent not found, we are working on a resolution
Some agents may timeout and be slow to load messages - we are actively investigating [Resolved]
Some users may be temporarily unable to access their agents. Service will be fully restored shortly.
There are potentially some service interruptions due to a failure of a core provider, service should stabalize soon
Resolved!
Investigating reports of elevated 409 (conversation is busy) causing degraded services
Conversations are being marked as busy, high latency - we're looking into it
Users using non-main agent conversations may experience 409/service related slowdowns and high latency. Users are advises to swap into the main or default conversation of their agent to unblock themselves in the meantime. We are looking into the cause.
Conversation status bug should be resolved
Some PRO users may have experienced an issue with rate limits, claiming they are out of credits, but were not. This issue occurred between 12:25 to 12:35 PM PST. Incident has been resolved
Some users are reporting steps created after the initial step are getting unintentionally rate limited by our system. We are working on a patch for this
Outage has been resolved