Inference API errors on all providers — credits lost, sandbox gone, no support response

[JustinGstar]

Hi Conway team,

I'm a user who set up an Automaton sandbox and funded it, but was unable to use the platform due to Inference API failures across all providers. I also emailed root@conway.tech twice (Feb 19 and a follow-up) with no response.

What happened

1. Created a Medium tier sandbox automaton-trader, us-east-1, 1 vCPU, 1GB RAM, 10GB disk)

2. Cloned and built the Automaton repo successfully

3. Ran the setup wizard — wallet created, API key provisioned, genesis prompt configured

4. Funded the agent with $17 in Conway credits (via multiple $5 USDC top-ups through x402)

5. Agent started correctly, detected credits, state repo initialized

The problem

Every model provider returned upstream errors when the agent tried to call the Inference API:

• OpenAI (gpt-5-nano, gpt-5-mini, gpt-4.1-mini): 429 - "You exceeded your current quota" — this is Conway's upstream OpenAI key, not mine

• Anthropic (claude-sonnet-4.5, claude-haiku-4.5): 400 - "Your credit balance is too low to access the Anthropic API" — Conway's Anthropic account, not my credits

• Kimi (kimi-k2.5): 401 - "Invalid Authentication" — Conway's Moonshot API key appears invalid

• Gemini: Listed as available: false on /v1/models

The /v1/models endpoint returned all models fine, and /v1/credits/balance showed $17 correctly. But no model actually worked.

Current state

• Sandbox ID: 6b151fc7d82cc2461aa683cd12376126

• Agent wallet: 0x3FFbb175646A71BD9f50265c09e142Ca0d86aEbd

• Conway account wallet: 0xF6aeb9472e522993670eBa8198d476557C1c160C

• Credit balance: $0.00 (the $17 was transferred to the agent wallet during setup, sandbox no longer exists)

• Wallet USDC (Base): $5.00

• VM creation failed 3 times with "worker error creating VM" before succeeding on the 4th attempt

What I'm asking for

1. Credit refund or restoration — I spent $17 on a service that didn't work due to platform-side inference errors

2. Status update on inference — Are the upstream API issues resolved? Is inference working now?

3. Support channel — Email to root@conway.tech went unanswered twice. Is there a better way to get support?

Happy to provide logs or any additional information.

Thanks,

Atamert Peker

[RealistSec]

same story here, all AI API providers failed.