Provider Operations
This is the Router operations board for provider keys, adapter blockers, and gateway smoke readiness. AgentCosts stays observe-first: providers can be tested through the gateway only after server-side keys exist, and automated routing stays gated behind trustworthy cost reports and budget alerts.
Readiness buckets
Gateway-supported providers: 9. These are the only candidates that should appear in observe-first proxy tests until native adapters are intentionally built.
Operating policy
Normalize OpenAI-compatible providers first: OpenAI, DeepSeek, Groq, OpenRouter, Kimi, Z.AI GLM, and MiniMax.
Keep native adapters for Anthropic and Gemini planned until the logging/reporting loop is stable.
Keep ROUTER_INGEST_API_KEY separate from provider API keys; it authenticates AgentCosts gateway ingest only.
Current rule
AgentCosts manages provider integrations as a non-secret catalog plus server-only environment-variable readiness. During beta, the Router observes and reports usage first; proxy routing should graduate provider by provider after cost reports and budget alerts are trusted.
Next integration queue
#1 / cheap / openai-compatible
Run a small observe-first gateway smoke test for DeepSeek R1, then review D1 event quality before expanding traffic.
Blocker: No credential blocker detected.
#2 / cheap / openai-compatible
Run a small observe-first gateway smoke test for DeepSeek V3.2, then review D1 event quality before expanding traffic.
Blocker: No credential blocker detected.
#3 / cheap / openai-compatible
Add MOONSHOT_API_KEY in Vercel Production before running Moonshot Kimi gateway smoke tests.
Blocker: Missing server environment variable: MOONSHOT_API_KEY.
#4 / cheap / openai-compatible
Add ZAI_API_KEY in Vercel Production before running Z.AI GLM gateway smoke tests.
Blocker: Missing server environment variable: ZAI_API_KEY.
#5 / fast / openai-compatible
Add GROQ_API_KEY in Vercel Production before running Groq gateway smoke tests.
Blocker: Missing server environment variable: GROQ_API_KEY.
#6 / fast / openai-compatible
Add MINIMAX_API_KEY in Vercel Production before running MiniMax gateway smoke tests.
Blocker: Missing server environment variable: MINIMAX_API_KEY.
#7 / premium / openai-compatible
Add OPENAI_API_KEY in Vercel Production before running DALL-E 3 gateway smoke tests.
Blocker: Missing server environment variable: OPENAI_API_KEY.
#8 / premium / openai-compatible
Add OPENAI_API_KEY in Vercel Production before running OpenAI o1 gateway smoke tests.
Blocker: Missing server environment variable: OPENAI_API_KEY.
cheap / openai-compatible
Credential
Server env
Gateway
Supported
Cheap lane for extraction, classification, and routine automation.
Models tracked
deepseek-v4-flash
Input $0.14/1M, cached input $0.0028/1M, output $0.28/1M
deepseek-v4-pro
Input $0.435/1M, cached input $0.003625/1M, output $0.87/1M
Next action
Run a small observe-first gateway smoke test for DeepSeek R1, then review D1 event quality before expanding traffic.
Env: DEEPSEEK_API_KEY
cheap / openai-compatible
Credential
Server env
Gateway
Supported
Cheap lane for extraction, classification, and routine automation.
Models tracked
deepseek-v4-flash
Input $0.14/1M, cached input $0.0028/1M, output $0.28/1M
deepseek-v4-pro
Input $0.435/1M, cached input $0.003625/1M, output $0.87/1M
Next action
Run a small observe-first gateway smoke test for DeepSeek V3.2, then review D1 event quality before expanding traffic.
Env: DEEPSEEK_API_KEY
cheap / openai-compatible
Credential
Server env
Gateway
Supported
Cost-efficient reasoning and coding alternative for agent workflows.
Models tracked
kimi-k2.6
Input $0.95/1M, cached input $0.16/1M, output $4/1M
kimi-k2.5
Input $0.6/1M, cached input $0.1/1M, output $3/1M
Next action
Add MOONSHOT_API_KEY in Vercel Production before running Moonshot Kimi gateway smoke tests.
Env: MOONSHOT_API_KEY
cheap / openai-compatible
Credential
Server env
Gateway
Supported
Low-cost GLM lane for Chinese and general automation workloads.
Models tracked
glm-5.1
Input $1.4/1M, cached input $0.26/1M, output $4.4/1M
glm-5
Input $1/1M, cached input $0.2/1M, output $3.2/1M
glm-4.7-flashx
Input $0.07/1M, cached input $0.01/1M, output $0.4/1M
Next action
Add ZAI_API_KEY in Vercel Production before running Z.AI GLM gateway smoke tests.
Env: ZAI_API_KEY
fast / openai-compatible
Credential
Server env
Gateway
Supported
Fast lane for realtime UX and short responses.
Models tracked
gpt-oss-20b
Input $0.075/1M, cached input $0.0375/1M, output $0.3/1M
llama-4-scout
Input $0.11/1M, output $0.34/1M
Next action
Add GROQ_API_KEY in Vercel Production before running Groq gateway smoke tests.
Env: GROQ_API_KEY
fast / openai-compatible
Credential
Server env
Gateway
Supported
Agentic coding and high-throughput text tasks where cost and speed both matter.
Models tracked
MiniMax-M2.7
Input $0.3/1M, cached input $0.06/1M, output $1.2/1M
MiniMax-M2.7-highspeed
Input $0.6/1M, cached input $0.06/1M, output $2.4/1M
MiniMax-M2.5
Input $0.3/1M, cached input $0.03/1M, output $1.2/1M
Next action
Add MINIMAX_API_KEY in Vercel Production before running MiniMax gateway smoke tests.
Env: MINIMAX_API_KEY
premium / openai-compatible
Credential
Server env
Gateway
Supported
Premium reasoning, coding, and baseline cost comparison.
Models tracked
gpt-5.5
Input $5/1M, cached input $0.5/1M, output $30/1M
gpt-5.5-pro
Input $30/1M, output $180/1M
gpt-5.4
Input $2.5/1M, cached input $0.25/1M, output $15/1M
+ 3 more models in provider JSON
Next action
Add OPENAI_API_KEY in Vercel Production before running DALL-E 3 gateway smoke tests.
Env: OPENAI_API_KEY
premium / openai-compatible
Credential
Server env
Gateway
Supported
Premium reasoning, coding, and baseline cost comparison.
Models tracked
gpt-5.5
Input $5/1M, cached input $0.5/1M, output $30/1M
gpt-5.5-pro
Input $30/1M, output $180/1M
gpt-5.4
Input $2.5/1M, cached input $0.25/1M, output $15/1M
+ 3 more models in provider JSON
Next action
Add OPENAI_API_KEY in Vercel Production before running OpenAI o1 gateway smoke tests.
Env: OPENAI_API_KEY
fallback / aggregator
Credential
Server env
Gateway
Supported
Fallback and long-tail model access during routing tests.
Models tracked
deepseek/deepseek-chat
Provider pricing
Next action
Add OPENROUTER_API_KEY in Vercel Production before running OpenRouter gateway smoke tests.
Env: OPENROUTER_API_KEY
not_routable / not-applicable
Credential
Catalog only
Gateway
Blocked
Catalog-only provider. Useful for discovery today; not yet part of Router logging or routing lanes.
Models tracked
E2B Sandbox
$0.0001/sec
Next action
Keep this provider in the public catalog until beta users ask for workflow-level cost tracking.
Env: not assigned
not_routable / not-applicable
Credential
Catalog only
Gateway
Blocked
Catalog-only provider. Useful for discovery today; not yet part of Router logging or routing lanes.
Models tracked
ElevenLabs
$0.30/1k
Next action
Keep this provider in the public catalog until beta users ask for workflow-level cost tracking.
Env: not assigned
not_routable / not-applicable
Credential
Catalog only
Gateway
Blocked
Catalog-only provider. Useful for discovery today; not yet part of Router logging or routing lanes.
Models tracked
Flux.1 [pro]
$0.055/img
Next action
Keep this provider in the public catalog until beta users ask for workflow-level cost tracking.
Env: not assigned
not_routable / not-applicable
Credential
Catalog only
Gateway
Blocked
Catalog-only provider. Useful for discovery today; not yet part of Router logging or routing lanes.
Models tracked
Piston API
$0.005/run
Next action
Keep this provider in the public catalog until beta users ask for workflow-level cost tracking.
Env: not assigned
not_routable / not-applicable
Credential
Catalog only
Gateway
Blocked
Catalog-only provider. Useful for discovery today; not yet part of Router logging or routing lanes.
Models tracked
SyncLabs
$0.15/min
Next action
Keep this provider in the public catalog until beta users ask for workflow-level cost tracking.
Env: not assigned
premium / native-adapter
Credential
Planned server env
Gateway
Blocked
Premium synthesis and long-context workflows.
Models tracked
claude-opus-4.7
Input $5/1M, cached input $0.5/1M, output $25/1M
claude-opus-4.6
Input $5/1M, cached input $0.5/1M, output $25/1M
claude-sonnet-4.6
Input $3/1M, cached input $0.3/1M, output $15/1M
+ 1 more models in provider JSON
Next action
Defer adapter work until observe-first OpenAI-compatible providers prove the savings report loop.
Env: ANTHROPIC_API_KEY
premium / native-adapter
Credential
Planned server env
Gateway
Blocked
Long-context and multimodal experiments.
Models tracked
gemini-3.1-pro
Input $2/1M, output $12/1M
gemini-3-flash-preview
Input $0.5/1M, cached input $0.05/1M, output $3/1M
gemini-3.1-flash-lite-preview
Input $0.25/1M, cached input $0.025/1M, output $1.5/1M
+ 3 more models in provider JSON
Next action
Defer adapter work until observe-first OpenAI-compatible providers prove the savings report loop.
Env: GOOGLE_GENERATIVE_AI_API_KEY
Catalog demand signals
These providers remain useful for acquisition and research, but they should not consume adapter work until a beta workflow proves repeated demand.