What your coding agent costs, how it caps you, and what changed this week.
15 agentic coding tools · updated 2026-10-22 · every entry links to its source.
- CLI 14
- IDE 4
- Self-hosted 1
Hybrid tools count as both CLI and IDE.
- limit 1
- model 0
- price 0
- feature 0
2026-10-21 to 2026-10-27 (build date)
Enforce a 5-hour rolling window plus a weekly cap on a paid plan: Claude Code, OpenAI Codex (CLI), OpenCode, Antigravity CLI (ex-Gemini CLI), Cline.
Derived from published plan limit text.
GPT-6.1 Sol (medium), AA Coding Agent Index v1.5. source retrieved 2026-10-05
Changelog
Newest first. Left border shows the kind; rose marks cuts and deprecations.
-
Claude 4.6 models and GPT-OSS-120b to be removed from the free Individual tier (announced). source
-
OpenCode Go's limited-time 4x boost on Step 5 Preview has ended: its monthly limit is back to $15 on Go and $60 on Go Plus (was $60 and $240). The 'Use balance' fallback is now called Extra usage. source
-
Kiro CLI 2.33.0 adds workspace trust: interactive V3 sessions ask before loading a folder's hooks, agents, MCP servers, steering or workflows. Model fallback is back in V3 sessions. source
-
Copilot deprecates Gemini 3.7 Flash, GPT-5.5, GPT-5.4, GPT-5.4 mini, GPT-5 mini and Grok 4.5 (announced Sep 18). source
-
OpenClaw 2026.10.5 adds Gemini 4 Argon (Google and Vertex) and Gemini 3.8 Flash; rerunning Google API-key setup picks Argon as the primary model. source
-
OpenCode Go drops Grok 4.6, Kimi K2.6 and MiniMax M2.7; Step 5 Preview no longer free on Go ($15 monthly allowance, temporarily $60). source
-
Cline extension 4.1.25: with no model chosen, the Cline provider uses Gemini 4 Argon instead of Claude Sonnet 5.5; OpenRouter stays on Sonnet 5.5. source
-
Gemini 4 Argon rolled out to paid Google AI Pro subscribers with a separate weekly-only Argon quota on top of the 3.8 Flash bucket. source
-
Copilot will retire Claude Opus 4.8, Claude Opus 5 and GPT-5.6 Terra on Nov 16 (Opus 4.8 stays for individual annual plans). source
-
Kiro CLI 2.32.0: folder-wide write approvals in V3 sessions, V3 cloud sessions runnable from non-interactive commands, and workflow plan previews before any step runs. source
-
Claude Code 2.1.299 waits 300s instead of 180s for a stream through a custom base URL. source
-
GPT-5.5 retired from Codex with ChatGPT sign-in (API-key sessions unaffected). source
-
Gemini 4 Argon added to Copilot Pro+, Max, Business and Enterprise at provider list price; not on Pro. source
-
With no model chosen, Cline and OpenRouter providers default to Claude Sonnet 5.5. source
-
Pro ($20), Pro+ and old Ultra closed to new signups. New plans via The X Subscription: Plus $8 (Grok only), Premium $30, Super $100, Ultra $200, with weekly (not monthly) allowances. Existing subscribers keep their plan until they switch. source
-
OpenClaw 2026.10.2 makes provider-managed compaction the default for eligible direct OpenAI Responses and Anthropic API-key chats. source
-
Kiro CLI 2.30.0 starts a gradual prompt to switch to the V3 harness; Kiro says V3 becomes the default with CLI 3.0 at the end of October. source
-
Grok Build bundled into The X Subscription (Plus $8 / Premium $30 / Super $100 / Ultra $200) with a weekly usage pool shared with Grok and Cursor. source
-
OpenCode 1.19.0: updates are no longer installed automatically (use /update), and `opencode upgrade --major` moves to OpenCode 2. source
-
Gemini Code Assist Standard and Enterprise no longer sold to new customers; existing annual plans can renew until Jan 31, 2027 and monthly until Dec 31, 2027. Google points customers to Antigravity via Gemini Enterprise. source
-
Ultrafast mode for GPT-6.1 Sol: Pro $500 only in Codex, uses 8x included usage (6x on purchased credits). source
-
Step 5 Preview made free on OpenCode for one week (promo ended Oct 16). source
-
OpenClaw 2026.9.9 points the `haiku` alias and Anthropic utility-model default at Haiku 5.5 (costs more on prompts over 100K). source
-
Max and Team plans now include monthly API credits ($100 Max 5x, $200 Max 20x); plan prices and usage limits unchanged. source
-
Claude Code 2.1.293 makes Haiku 5.5 the default Haiku on the API ($0.10/$0.50 per 1M tokens; $0.50/$2.50 on prompts over 100K). source
-
Codex CLI 0.161.0 makes GPT-6.1 Sol the default model. source
-
OpenAI API usage tiers cut from five to three (Build, Launch, Grow: $500 / $5,000 / $200,000 monthly usage limits; Grow now needs $500 in total payments); affects Codex on an API key. source
-
Copilot retires Gemini 3.5 Flash, Gemini 3.6 Flash, Kimi K2.7 Code and Claude Opus 4.7. source
-
Claude Code 2.1.287 defaults Opus 4.7+ and Fable to a 1M-token context on Bedrock, Vertex, Foundry and the Claude apps gateway; CLAUDE_CODE_DISABLE_1M_CONTEXT=1 keeps 200K. source
-
Business and Enterprise seats paid by card/PayPal now charged upfront; prices unchanged. source
-
Pro 500 ($500) launched with Ultrafast and 25x Plus usage. Pro 200 reopened to new subscribers at 10x Plus Codex/Work usage (was 20x); grandfathered Pro 200 subscribers keep 20x through Oct 29. source
-
New Go Plus plan at $40/month with higher per-model limits; Go stays $10. source
-
ClinePass drops GLM-5.2, Kimi K2.6, Kimi K2.7 Code and DeepSeek V4 Flash; adds DeepSeek V4.1 Flash and Muse Spark 1.3. Price unchanged. source (via tracker)
-
Cline ends the $4.99 ClinePass first-month offer. source (via tracker)
-
Gemini CLI free tier ended; free users moved to Antigravity CLI with a weekly quota. source
-
Usage-based billing (GitHub AI Credits) live on all Copilot plans; Copilot Max launched as an upgrade. source
-
Qwen OAuth free tier discontinued (cut from 1,000 to 100 requests/day on Apr 13, then ended). source
Will You Hit the Cap? (Quick BurnRate Estimator)
Rough estimate based on published limits, not a guarantee.
Daily Builder (4-6h/day)
PreferenceHow the estimate works (per-tool cap_risk)
cap_risk rates each tool's ~$20 plan (the paid plan between $8 and $30 closest to $20) from its published limit text: high = 5-hour rolling window plus weekly limit, medium = 5-hour window only, low = no 5-hour window. The warning shows for high at Daily Builder and above, and for medium at Heavy. Best fit = matching tools without a warning. Tools with a paid subscription plan rank ahead of BYOK-only tools (except under "Must allow BYOK", which leads with uncapped BYOK tools); then by cap strictness (rolling windows before monthly credit pools), then sub-agent support at Heavy, then benchmark cost per task, then cheapest paid plan. "Fixed monthly bill" = has a priced subscription plan (BYOK-only / API-billed tools excluded). "Terminal (CLI)" = CLI, terminal, hybrid or self-hosted, or a documented headless command.
| Tool | ~$20 plan | cap_risk | Why |
|---|---|---|---|
| Antigravity CLI (ex-Gemini CLI) | Google AI Pro $20 | high | Google AI Pro $20: 5-hour rolling window plus weekly limit |
| Claude Code | Pro $20 | high | Pro $20: 5-hour rolling window plus weekly limit |
| Cline | ClinePass $9.99 | high | ClinePass $9.99: 5-hour rolling window plus weekly limit |
| OpenAI Codex (CLI) | ChatGPT Plus $20 | high | ChatGPT Plus $20: 5-hour rolling window plus weekly limit |
| OpenCode | Go $10 | high | Go $10: 5-hour rolling window plus weekly limit |
| Kimi Code CLI | Plus $19 | medium | Plus $19: 5-hour rolling window |
| Cursor | Premium $30 | low | Premium $30: no 5-hour window in published limits |
| GitHub Copilot CLI | Pro $10 | low | Pro $10: no 5-hour window in published limits |
| Grok Build | Premium $30 | low | Premium $30: no 5-hour window in published limits |
| Kiro CLI | Pro $20 | low | Pro $20: no 5-hour window in published limits |
| Warp | Build $20 | low | Build $20: no 5-hour window in published limits |
| Aider | — | n/a | No paid plan between $8 and $30 |
| Goose | — | n/a | No paid plan between $8 and $30 |
| OpenClaw | — | n/a | No paid plan between $8 and $30 |
| Qwen Code | — | n/a | No paid plan between $8 and $30 |
Compare all 15 tools
Search, filter, sort, and tick 2-3 tools to compare side by side.
| Compare | Models | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Aider | cli | yes | yes | no | none documented | free / BYOK | BYOK, no vendor cap | Any provider via LiteLLM |
— | — | ||
| Antigravity CLI (ex-Gemini CLI) reports | hybrid | ? | ? | ? | unknown | $20 | 5h + weekly rolling | Gemini 4 Argon (paid) Gemini 3.8 Flash +2 more |
? | $5.84 | 2026-10-15 | |
| Claude Code | cli | no | yes | yes | client | $20 | 5h + weekly rolling | Claude (Fable, Opus, Sonnet, Haiku) |
$3.33 | 2026-10-14 | ||
| Cline | hybrid | yes | yes | yes | client | $9.99 | 5h + weekly rolling | Any provider; ClinePass: DeepSeek V4.1 Flash, GLM-5.3, Kimi K3, Muse Spark 1.3, others |
? | — | 2026-10-16 | |
| Cursor reports | ide | no | ? | ? | client | $8 | Weekly rolling | Grok 4.7 Grok 4.6 +3 more |
? | — | 2026-10-12 | |
| GitHub Copilot CLI | cli | no | ? | yes | client | $10 | Monthly credit pool | Multi-vendor (Claude, GPT, Gemini, Kimi) |
— | 2026-10-19 | ||
| Goose | cli | yes | yes | yes | client | free / BYOK | BYOK, no vendor cap | Any provider via config |
— | — | ||
| Grok Build | cli | no | yes | yes | client | $8 | Weekly rolling | Grok 4.7 grok-build-0.1 |
$8.82 | 2026-10-12 | ||
| Kimi Code CLI | cli | ? | ? | ? | unknown | $19 | Monthly credit pool | Kimi K3 |
? | $5.05 | — | |
| Kiro CLI | cli | no | ? | yes | client | $20 | Monthly credit pool | ? | — | 2026-10-20 | ||
| OpenAI Codex (CLI) reports | hybrid | yes | yes | yes | client+server | $8 | 5h + weekly rolling | GPT-6.1 Sol GPT-6 Astra +2 more |
$0.705 | 2026-10-14 | ||
| OpenClaw | self-hosted agent | yes | yes | yes | unknown | free / BYOK | BYOK, no vendor cap | Anthropic, OpenAI, Gemini, MiniMax, OpenRouter, local; Claude/Codex harnesses as plugins |
? | — | 2026-10-18 | |
| OpenCode | cli | yes | yes | yes | client | $10 | 5h + weekly rolling | 75+ providers incl. local; Go: curated open models |
$4.24 | 2026-10-22 | ||
| Qwen Code | cli | yes | yes | yes | client | free / BYOK | BYOK, no vendor cap | Qwen models; other OpenAI-compatible providers |
— | 2026-04-15 | ||
| Warp | terminal | no | yes | no | client | $20 | Monthly credit pool | Multi-vendor |
? | — | — |
Tick 2-3 tools to compare side by side. Cap strictness is derived from published plan limit text: green = BYOK-only with no vendor plan cap, amber = rolling 5h/weekly windows, red = monthly credit pool that cuts off. *Cost per task: Artificial Analysis Coding Agent Index v1.5 via aicharts.io, one config per tool, retrieved 2026-10-05; not comparable across very different configs. "?" = not verified yet.