I am a heavy user of agentic coding tools, part of my job is to evaluate their capabilities and limitations. This is an experiment where I am tracking their capabilities, limitations and my overall impression of them.

Note: Some cells have more information when you hover over them.

Colour Key:

  • Requires subscription
  • Minor limitation
  • Moderate
  • Significant
  • Major
Claude CodeOpenCodeOpenAI CodexClineGithub CopilotGoogle AntigravityCursorAmazon Kiro
Functionality
Inference Providers & Model SelectionAny (works best with Anthropic models)AnyOpenAI models only, US based sub or BYO baseURLAnyUS based sub or BYO baseURLUS based Gemini models onlyUS based models, limited by plan, BYO option limits featuresUS based sub or BYO baseURL
Sub-AgentsYesYesYesYesNoPartial / very limitedYes (limited)No
Custom AgentsYesYesNoYesYesYesNoNo
Prompt templatesYesYesNoYesYesPartialYesNo
Coding abilitiesBestGoodGoodVery goodLower endMediumMid rangeMid range
SpeedFastestGoodGoodFastestSlowMediumSlowSlow
MCP Spec alignmentHighMedium/HighMedium/HighMedium/HighHighMediumMediumLow
Full model capabilityFullFullFullFullHeavily Context LimitedFullContext LimitedContext Limited
Operational management
Consumption budget controlsWith EnterpriseNoPartial (basic project/key limits)With EnterpriseN/ANoNo (for BYO)N/A
Developer experience
User InterfaceVSCode, IntelliJ, TUI, Zed, CLI, CI/CD, Web, DesktopTUI, CLIVSCode IDE, CLIVSCode IDE, IntelliJ, CLIVSCode, Zed, CLIStandalone IDE (Fork of VSCode that adds no value), CLIStandalone IDE (Fork of VSCode), CLIStandalone IDE (Fork of VSCode that adds no value), CLI
SourceClosed sourceOpen sourceOpen sourceOpen sourceOpen sourceClosed sourceClosed sourceClosed source
Babysitting requiredVery LowMedium/LowMedium/LowLowVery HighHighVery HighVery High
Plan/Act modesYesYesYesYesYesPartial (doesn't work)YesNo
Workflow flexibilityHighHighHighHighMedium/LowMedium/LowMedium/LowVery Low
Community sizeVery LargeMediumMediumMediumLargeSmallLargeSmall
Configuration management
File based configurationYesYesYesNoYesLimitedNoLimited - no RBAC or rules
Per-project configuration (rules, MCP, RBAC)YesYesPartialIn betaPartialLimitedNoLimited - no RBAC
Centralised configurationPartial (manual - split across config files)NoNoWith enterprise / teams subscriptionPartialNoNo (for BYO), Yes for US basedNo
Automation
Automation hooksYesYes (Plugins)YesYesNoNoNoYes
Headless / CLI / CI/CD modeYesYesYesYesYesYesIn betaNo
Automated Context CleanupYesYes (Plugin)YesNoYesNoNo
Access control
RBAC - CommandsYesYesPartialPartial (basic)NoNo (has basic allow list - easy to bypass)No (has basic allow list - easy to bypass)No (has basic allow list - easy to bypass)
RBAC - Web domainsYesYesNoWith enterprise / teams subscriptionNoNoNoNo
RBAC - MCP toolsYesYesNoWith enterprise / teams subscriptionNoNoNoNo
SSOYesN/ANoWith enterprise / teams subscriptionYesYesNo (for BYO), Yes for US basedYes
User Interface (Subjective)
User InterfaceBestGoodGoodBestGoodGoodGoodAverage
Cost (Estimate)
Monthly cost ($USD) estimate for agentic coding (used to the best of the tools limitations)$100-$200$200-$4000 (depends on sub or consumption)$200$200-$4000 (depends on sub or consumption)$100-$300$200+$200+$200++