HERMES NODE
>_
SnipGeekLABS
Back to Research Feed
AI Stack Live Registry

Providers & Gateway Model Registry

Direct inventory of the current coding stack, provider history, and API plan economics. ChatGPT Codex Trial is the primary workflow, with two trial windows covered through 3 Nov 2026.

Primary stack status

ChatGPT Codex Trial

2 trial windows active: ends 18 Oct 2026 · ends 3 Nov 2026, evaluated in Asia/Jakarta time.

Decision needed

The two windows rotate: when one hits its limit, work moves to the other. Telegram reminders are scheduled for 15, 16, and 17 October. The trial does not auto-convert after it ends.

Recommended providers

CommandCode Go (GOAT quota full)

Current Provider API route · $1.36/mo Go; GOAT plan $10/mo (100% used)

Active

Antigravity AI

Secondary AI agent · Free tier through May 2027

Active

Not recommended

Kiro AI

Retired

Trial reverted early and the connection was unreliable for coding work.

ClinePass

Retired

Model availability and coding stability did not justify continuing the subscription.

DeepInfra

Retired

The prepay model was not cost-effective for this workflow.

All providers & subscriptions

ProviderStatusCostVerdict
ChatGPT Codex TrialPrimary coding agentDecision neededTrialspent $0Recommended
CommandCode Go (GOAT quota full)Current Provider API routeActive$1.36/mo Go; GOAT plan $10/mo (100% used)spent ~$16.32Recommended
Antigravity AISecondary AI agentActiveFree tier through May 2027spent $0Recommended
Token Harbor Agent PassAPI gateway and paid usage allowanceActive$0.99 first month; then $1.99/mospent $0.99Utility
B.AI Free ModelsFormer free gatewayRetired$0 while availablespent $0Utility
DeepSeek API (Direct)Former direct APIRetired$1.70 exhaustedspent $9.54Utility
Kiro AIFormer trial providerRetiredFree, limited usable creditsspent $0Not recommended
ClinePassFormer API proxyRetired$10.60/mospent $5.30Not recommended
DeepInfraFormer inference providerRetired$5 prepayspent $5.00Not recommended
OpenCode GoFormer coding agentRetired$5 first monthspent $5.00Utility

Provider status reviewed 6 Oct 2026. The Codex trial is evaluated in Asia/Jakarta time.

9Router model ID collection

Current gateway routes are documented here for fallback planning. Retired providers remain visible only as history.

Current model route

Use ChatGPT Codex for the primary coding workflow. CommandCode Go is the current Provider API route; the GOAT allowance is 100% used and appears below for plan-value reference only. Token Harbor Agent Pass is the paid API backup, while its free-model allowance remains separate. Antigravity is the secondary agent route. Desktop and server agents (Reasonix, Hermes Agent, OpenCode, Zed IDE) reuse these gateways and keys rather than adding separate subscriptions.

cmc/ CommandCode

cmc/deepseek/deepseek-v4-procmc/deepseek/deepseek-v4.1-flashcmc/MiniMaxAI/MiniMax-M3cmc/z-ai/glm-5.3-flashcmc/xiaomi/mimo-v2.5cmc/stealth/space-bunny-alpha

Archived routes

B.AI was retired on 16 Sep 2026. Its former route is not included in the current 9Router recommendation.

CommandCode plan-value tracker

Updated 21 Oct 2026 22:23 WIB

CommandCode plan economics for comparison; current usage is on Go. The GOAT allowance has reached 100% used, so the GOAT figures below are reference only, not available balance. Data is sourced from CommandCode documentation .

GOAT plan-value reference

List price 10/mo; listed allowance 70 (currently 100% used)
1.Jev595,238 chats/mo
2.DeepSeek V4.1 Flash (latest)Vision205,128 chats/mo
3.DeepSeek V4 Flash (latest)153,846 chats/mo
4.MiMo V2.5Vision97,403 chats/mo
5.Muse Spark 1.3 ContributorVision90,909 chats/mo

Top model value

1.Muse Spark 1.2 Contributor1 credit: 4,200 chats
2.Muse Spark 1.3 Contributor1 credit: 4,200 chats
3.MiMo V2.5deal 98%1 credit: 3,000 chats
4.MiMo V2.6 Flash1 credit: 3,000 chats
5.DeepSeek V4 Flash1 credit: 2,400 chats
6.DeepSeek V4 Flash Vision (exp)1 credit: 2,400 chats
7.DeepSeek V4.1 Flash1 credit: 2,400 chats

Free tier

Beam 1.0
while capacity lasts
Glyph Cluster
while the stealth preview lasts
Laguna S 2.1
while capacity lasts
MiniMax M3.1 Flash
for a limited time
Ornith 1.5 122B
until 11:00 PM PT on October 21

Active deals

MiMo V2.5 Pro99%
MiMo V2.598%
MiniMax M350%

DeepSeek window

OFF-PEAK

08–11 & 13–17 WIB (Mon-Fri)

Changes at 08:00 WIB

Separate free API routes

Token Harbor free routesSeparate from the paid Agent Passtokenharbor.ai

The free-model allowance continues alongside the Agent Pass and follows its own reset schedule. Current free models shown in the tracker: MiMo V2.6 Flash, DeepSeek V4.1 Flash, and TH-Rudder.

Token Harbor API value watch

Checked 2026-10-21T22:23:06.645972+07:00

Published per-token rates in USD per 1M tokens. Sources: Value catalog , pricing , and billing docs .

Agent Pass

$0.99 first month, then $1.99/month

$10 shared usage; selected model boosts can stretch its value to $20. Allowance refreshes in 4 weekly windows of 7 days; 5% off PAYG after the allowance.

Adds: DeepSeek V4.1 Flash 1017 / Claude Haiku 5.5 / GLM 5.3 Flash. Free allowance remains separate: DeepSeek V4.1 Flash / MiMo V2.6 Flash / TH-Rudder.

DeepSeek on Token Harbor

Discount shown: 21:00–07:00 WIB (+1 day)

$0.04 input / $0.15 output per 1M during that window.

The catalog does not state the applicable days. DeepSeek direct API has its own schedule: weekends are off-peak all day; do not transfer that rule to Token Harbor.

ModelInput / 1MOutput / 1MListed discount
GPT-6 Luna$0.10$0.50—
GLM 5.3 Flash$0.15$0.50—
Qwen3.8 Flash$0.15$0.47—
DeepSeek V4.1 Flash$0.07$0.30$0.04 / $0.15

Cost-only guidance for this set: DeepSeek V4.1 Flash has the lowest input rate; DeepSeek V4.1 Flash has the lowest output rate. GPT-6 Luna is a reasonable default for input-heavy routine calls, not a quality ranking.

How I wire everything

ChatGPT Codex is the primary coding workflow during the trial. Two trial windows run in parallel and rotate on limit — when one is rate-limited, work moves to the other — covering the stack through 3 Nov 2026. CommandCode Go is the current Provider API route; GOAT is not being used because its allowance has reached 100%.

Antigravity remains the secondary agent route. Token Harbor Agent Pass started on 6 Oct 2026 at $0.99 for the first month, then $1.99/month if continued. It provides $10 of shared usage value (selected boosts can stretch this to $20) and 5% off PAYG; its separate free-model allowance remains available. Token Harbor says the pass renews at $1.99 unless canceled; the account's auto-renew setting is not recorded here. B.AI was retired on 16 Sep 2026 and is no longer part of the active 9Router recommendation.

The agent clients in this stack — Reasonix, Hermes Agent, OpenCode, Zed IDE — are not separate providers or subscriptions. Reasonix, for example, uses the same CommandCode Go API, 9Router gateway, and Token Harbor routes listed above, so it adds no additional line item.

The server checks the Codex trial each morning and sends Telegram review reminders on 15, 16, and 17 October. Because a second window runs through 3 Nov 2026, nothing needs to change when the first window ends on 18 Oct 2026.

Active

3

Retired

6

Trial coverage

3 Nov 2026

GOAT allowance used

100%