VERDICT HISTORY
- Shipped Command Code flagship card — terminal agent with taste learning, open-model deals, and full pricing/privacy research from commandcode.ai.
Command Code88FlagshipBenchmark pendingTaste-aware terminal coding agent and open-model harnessVerified 2026-06-15
Dossier · Build & code
Command Code
Taste-aware terminal coding agent and open-model harness · last verified 2026-06-15
FlagshipEDITOR'S PICK
Build & code
Command Code
88/100
ROLETaste-aware terminal coding agent and open-model harness
Editorial fit
88
Source quality
71
Citation honesty
58
Privacy controls
78
Value for money
91
Speed
78
$1/MOGo
$15/MOPro
$100/MOMax
EDITOR'S NOTE
Desk hands-on 2026-06: The harness is the story — taste-1 learns how you actually ship (pnpm vs npm, vp-* components, finish-the-task discipline) and open models stop feeling like toys. Budget the credits; review the skills file like code.
AT A GLANCE
Terminal coding agent with continuous taste learning (taste-1), open-model deals, MCP/skills, and subscription credits from $1/mo — ships, tests, and refactors in your conventions.
Role: Taste-aware terminal coding agent and open-model harnessCategory: Build & codeEditorial · hands-on
Sixteen scored dimensions for the taste-aware terminal agent — credits, open models, .commandcode/ storage, and provider routing checked 2026-06-15.
PUBLIC FACTS · vendor & repo
List prices and pay-as-you-go entry points we can cite without running our own bench. Each tile links to a source when possible.
Limits & product surface
Non-price vendor claims — multipliers, caps, and API scope. Detailed matrices live in subscription and SDK sections below.
- Primary surface
- prosumer-saasCard identity ↗
- Taste storage
- .commandcode/ in repo (local-first)docs.commandcode.ai/taste ↗
- Modalities
- code, textCard identity ↗
- Evidence tier
- editorial-hands-onQuality gate ↗
HOW IT WORKS · agent loop
The public positioning for this product — the loop we score against on VerdictPal.
Install and taste bootstrap
Run cmd in repo root; accept/reject edits so .commandcode/taste accrues package-manager and style rules.
Pick model and autonomy
Match DeepSeek deals to boilerplate; reserve Opus for architecture-sensitive diffs.
Review skills file like code
Treat accumulated taste rules as versioned config — bad patterns compound silently.
METRIC LAB · 16 DIMENSIONS
Click a tile for the editorial note. Color follows score: coral, yellow, mint.
Shape
Avg 85 · 72–92
88Editorial fit
Terminal agent with taste learning — strong for repo-native builders, not literature search.
By group
- Editorial fit88
- Taste learning92
- Agent depth86
- Model breadth88
- Value for money91
- Credit clarity72
- No-training posture88
- Local taste storage90
- Diff discipline80
- Open-model harness90
- Skills & MCP84
- Headless automation85
- Team taste share82
- Reliability84
- Competitive position86
- Setup friction78
SEARCH MODES · 4 lenses
One input box, many retrieval postures. Filter by tier.
01Free
Interactive CLI
Default terminal session with taste learning on accept/reject/edit
02Free
Headless -p
Scripted runs for CI and batch refactors
03Free
Open-model deals
DeepSeek and gateway routes with credit stretch on Go plan
04Pro
Premium routes
Claude Opus and frontier models on Pro/Max credits
BENCHMARK LEDGER
Public rows are vendor or third-party claims we logged with a date. Desk rows are reserved for VerdictPal self-run results.
VerdictPal-run benches. Completed rows stay in the table; planned checks are collapsed below so unfinished work does not masquerade as evidence.
| Measure | Result | Status | Source |
|---|---|---|---|
| CLI feature ship: DeepSeek V4 Flash vs Cursor Agent on same repoTaste file accrued workflow preferences; open-model output matched premium on boilerplate. | Desk hands-on | completecommand-code-models-v1 | VerdictPal desk |
Planned desk checks (1)
- 10 headless -p tasks with vs without .commandcode/taste skillsMismatch rate on package manager, test runner, and CSS conventions before/after taste file.
PRICING DECK · checked 2026-06-15
Go
$1/mo
+ processing fee
- $10 credits (~$40 on DeepSeek deals)
- taste-1 access
Pro
$15/mo
Side projects
- $30 credits
- Open + premium models
RAW · checked 2026-06-15: Go $1/mo + processing fee ($10 credits, ~$40 effective on DeepSeek V4 Pro deals) · Pro $15/mo ($30 credits) · Max $100/mo · Ultra $200/mo · Provider $15/mo API · Teams $40/mo pooled · Enterprise custom · top-up credits roll over at API cost
RESEARCH LOG · desk notes
What we learned while building this dossier — not vendor copy.
Go plan and processing fee verified
$1/mo Go includes $10 credits plus processing fee on commandcode.ai/pricing.
COMPETITIVE LENS
Where Command Code wins for cited research — and where a rival still belongs in the stack.
Command Code wins when taste learning and open-model economics matter in the terminal; Cursor wins for IDE-native diff review and tab completion.
Command Code
- You want .commandcode/ skills that learn pnpm vs npm from edits
- You budget credits across DeepSeek deals instead of one IDE subscription
Cursor Agent
- You need inline IDE edits and LSP-native navigation
- Your team standardizes on Cursor-only workflows
UNDER THE HOOD
Vendors named on the product about page — useful for procurement and privacy reviews.
- IDE companion
- Cursor
- Open agent alt
- OpenCode
- Terminal surface
- Warp
DEEP PANES · 6 LENSES
Editorial lenses only. Subscription and API pricing live in their own sections above.
Lens 01
Taste learning
Every accept, reject, and edit becomes a signal. taste-1 distills conventions into project skills — the repo already has a .commandcode/taste/taste.md from VerdictPal work.
- Desk signal
- Taste file exists in this repo
- Let taste accrue for a week before judging diff quality
- Review generated skills in PRs — delete bad patterns
TEST SCENARIOS · hands-on lab
How we exercised the product. Step through each run before you trust the scores.
Test run
Step 1 of 3
Install cmd; run interactive sessions on a real repo branch
BEST FOR
- Developers who want a terminal-native agent that learns package manager, test runner, and style preferences from edits
- Builders experimenting with DeepSeek, Qwen, Kimi, and other open models without writing a custom harness
- Solo devs and small teams who want $1/mo entry with real agent tooling (MCP, skills, headless -p)
- Repos already using .commandcode/ taste files that should stay portable via npx taste push/pull
AVOID IF
- You need a GUI IDE with Tab completion as the primary surface — use Cursor instead
- You cannot parse credit deals, processing fees, and per-model burn before approving team spend
- You will run headless agents on sensitive repos without tests, sandbox review, or deny lists
- You need guaranteed EU-only inference on every model route without reading gateway policies
STRENGTHS
- npm i -g command-code CLI (cmd)
- taste-1 continuous learning into .commandcode/ skills and /memory
- Interactive, headless (-p), and sandbox agent modes
- /skills, /commands, /mcp, hooks, plugins
- /agents and persistent session memory
- Built-in file ops, shell, grep, extended thinking
- Model picker: Claude, GPT, DeepSeek, Qwen, Kimi, GLM, MiniMax, BYOK
- Studio billing, usage analytics, team pooled credits
- npx taste push/pull and /share sessions
- Commits and PR flows from the terminal
WEAKNESSES
- Prompts and code context still leave the machine to third-party model providers when you invoke AI
- Credit math (deals + processing fees + top-ups) is easy to mis-budget for teams
- Taste can encode bad habits if you accept sloppy diffs — garbage in, skill out
- Headless --yolo mode can execute destructive shell without IDE-style apply review
- Open-model routes depend on gateways (Vercel, Cloudflare, OpenRouter) with shifting downstream hosts
- Younger product vs Cursor/Factory — fewer institutional procurement templates
HOW IT COULD IMPROVE
- Institutional policy friction could be reduced with a local/on-prem deployment option, or at minimum with clear data-handling documentation that compliance teams can review without an NDA.
- Institutional policy friction could be reduced with a local/on-prem deployment option, or at minimum with clear data-handling documentation that compliance teams can review without an NDA.
EDITORIAL EVIDENCE · 3 ENTRIES
TASK
Ship a small CLI feature with Command Code on DeepSeek V4 Flash vs Cursor Agent on the same repo fixture.
Desk hands-on on VerdictPal: taste file accrued workflow preferences; open-model output matched premium agents on boilerplate tasks.
↳ Time-to-PR, test pass rate, diff review burden, credit burn, model routing clarity.
TASK
Run 10 headless -p tasks with and without accumulated .commandcode/taste skills.
Pending formal bench; anecdotal reduction in package-manager and test-runner mismatches after one week.
↳ Skill file size, bad-pattern retention, sandbox vs --yolo safety.
TASK
Compare Go-plan credit burn across DeepSeek deal vs Claude Opus on identical refactor prompts.
Pending; pricing page claims up to 4× stretch on DeepSeek V4 Pro.
↳ Effective $/task, processing fee impact, top-up triggers.
PRIVACY DEEP-DIVE · checked 2026-06-15
Langbase, Inc. d/b/a Command Code states it does not train on your source code or sell personal data. Taste preference rules store locally under .commandcode/; optional account sync uploads structured taste profiles, not full repos. AI prompts, snippets, and attachments route to selected providers (Anthropic, OpenAI, Google, gateways, open-model hosts) under their terms. Billing via Stripe; hosting on US cloud with EU options noted for some enterprise routes.
Training on your dataOff by default
EU data residencyEnterprise tier only
SOC 2 attestationNot documented
Local-first by defaultLocal option available
ALTERNATIVES
Head-to-head
Compare side-by-side → open compare
WORKFLOW ROLES
How this tool fits into a composed research stack:
terminal coding agenttaste-aware refactorheadless agent scripts
QUALITY GATE · FLAGSHIP
Evidence ready
Benchmark pending
- desk hands-on 2026-06-15: active .commandcode/taste/taste.md in VerdictPal repo from real agent sessions.
- Public recommendation: true
SOURCE LEDGER · 5 checked
- Product sitechecked 2026-06-15
- Pricingchecked 2026-06-15
- Privacychecked 2026-06-15
- Docschecked 2026-06-15
- Taste docschecked 2026-06-15
Related dossiers
The Pack · Editorial newsletter
New cards in your inbox. Free.
One short email when a card ships or changes status. No tracking, no third-party analytics. Unsubscribe in one click.