# Feedback Bench

> Coding agents ranked by what their users say in public (Reddit, X, G2, Trustpilot). 17 agents, 63 criteria in 9 areas, built 2026-10-01, window 2026-08-31 to 2026-09-27. Every number comes from labelled public posts; every page lists its posts with links.

The web page renders with JavaScript. Every view below is also available as static Markdown and every number as JSON or CSV, so agents do not need to run the page. The Feedback Score is 100 × √(Popularity × Customer love). In the data files, popularity is `reach` and customer love is `regard`.

## Start here

- [Ranking, top quadrant and areas](https://feedbackbench.com/index.md): the landing page as Markdown
- [Method](https://feedbackbench.com/method.md): sources, labelling, formulas, parameters, weights, and all 63 criterion definitions
- [Everything in one file](https://feedbackbench.com/llms-full.txt): ranking, method, and every agent page without posts

## Agents

- [Claude Code](https://feedbackbench.com/agents/claude-code.md): rank 1, Feedback Score 70.4, Anthropic
- [OpenAI Codex](https://feedbackbench.com/agents/codex.md): rank 2, Feedback Score 67.4, OpenAI
- [OpenCode](https://feedbackbench.com/agents/opencode.md): rank 3, Feedback Score 66.0, Anomaly (open source)
- [Cursor](https://feedbackbench.com/agents/cursor.md): rank 4, Feedback Score 59.7, Anysphere
- [Devin](https://feedbackbench.com/agents/devin.md): rank =5, Feedback Score 55.9, Cognition
- [Google Antigravity](https://feedbackbench.com/agents/antigravity.md): rank =5, Feedback Score 54.9, Google
- [Pi](https://feedbackbench.com/agents/pi.md): rank 7, Feedback Score 52.7, Earendil Works (open source)
- [GitHub Copilot](https://feedbackbench.com/agents/copilot.md): rank =8, Feedback Score 43.0, GitHub
- [Cline](https://feedbackbench.com/agents/cline.md): rank =8, Feedback Score 42.9, Cline (open source)
- [Zed](https://feedbackbench.com/agents/zed.md): rank =8, Feedback Score 42.4, Zed Industries
- [Factory](https://feedbackbench.com/agents/factory.md): rank =11, Feedback Score 35.6, Factory
- [Amp](https://feedbackbench.com/agents/amp.md): rank =11, Feedback Score 35.3, Amp
- [Kiro](https://feedbackbench.com/agents/kiro.md): rank 13, Feedback Score 29.5, AWS
- [Conductor](https://feedbackbench.com/agents/conductor.md): rank 14, Feedback Score 26.4, Melty Labs
- [Warp](https://feedbackbench.com/agents/warp.md): rank 15, Feedback Score 23.2, Warp
- [Grok Build](https://feedbackbench.com/agents/grok-build.md): rank 16, Feedback Score 12.6, xAI
- [Augment Code](https://feedbackbench.com/agents/augment.md): rank 17, Feedback Score 9.7, Augment

## Areas and criteria

- [Paying and limits](https://feedbackbench.com/criteria/paying.md): What do you pay, and how far does it get you?
  - [How much use a plan's price buys](https://feedbackbench.com/criteria/limits.plan_value.md)
  - [Short rolling usage window blocks or interrupts work](https://feedbackbench.com/criteria/limits.window_interrupts_work.md)
  - [Single prompt, model or effort level consumes disproportionate quota](https://feedbackbench.com/criteria/limits.burn_rate.md)
  - [Price, allowance or plan terms changed](https://feedbackbench.com/criteria/limits.allowance_change.md)
  - [Quota reset timing and bonus or banked resets](https://feedbackbench.com/criteria/limits.reset_schedule.md)
  - [Usage meter visibility and accuracy](https://feedbackbench.com/criteria/limits.usage_meter.md)
  - [Prompt cache hits, misses and invalidation](https://feedbackbench.com/criteria/limits.prompt_cache.md)
  - [Pay-as-you-go overage, fallback billing and spend caps](https://feedbackbench.com/criteria/billing.overage_charges.md)
  - [Pricing and plan terms stated clearly and consistently](https://feedbackbench.com/criteria/billing.pricing_clarity.md)
  - [Free tier and free model availability and limits](https://feedbackbench.com/criteria/billing.free_tier.md)
  - [Using an existing subscription across tools](https://feedbackbench.com/criteria/billing.subscription_portability.md)
- [Setting up and connecting](https://feedbackbench.com/criteria/setup.md): How hard is it to install and connect?
  - [Install, launch and sign-in](https://feedbackbench.com/criteria/setup.install_signin.md)
  - [Connecting own API keys, local models and custom endpoints](https://feedbackbench.com/criteria/setup.provider_byok_local.md)
  - [MCP servers, plugins, skills and hooks](https://feedbackbench.com/criteria/setup.extensions_mcp.md)
  - [Onboarding, discoverability and documentation](https://feedbackbench.com/criteria/setup.onboarding_docs.md)
  - [IDE and editor integration](https://feedbackbench.com/criteria/setup.ide_integration.md)
- [Choosing models](https://feedbackbench.com/criteria/models.md): Which models do you get, and do they hold up?
  - [Which models are offered on a plan and when](https://feedbackbench.com/criteria/models.catalog_access.md)
  - [Automatic model routing and fallback](https://feedbackbench.com/criteria/models.routing_auto.md)
  - [Reasoning effort setting and its defaults](https://feedbackbench.com/criteria/models.effort_control.md)
  - [Quality got worse or better over time](https://feedbackbench.com/criteria/models.quality_drift.md)
- [Instructing and context](https://feedbackbench.com/criteria/context.md): Does it follow your instructions and keep the right context?
  - [Persistent project rules files are read and obeyed](https://feedbackbench.com/criteria/context.instruction_files.md)
  - [Direct in-prompt instructions and caps are followed](https://feedbackbench.com/criteria/context.instruction_following.md)
  - [Asks the user versus guessing](https://feedbackbench.com/criteria/context.clarifying_questions.md)
  - [Output degrades as the context window fills](https://feedbackbench.com/criteria/context.long_context_decay.md)
  - [Context compaction keeps what matters, cheaply and quickly](https://feedbackbench.com/criteria/context.compaction.md)
  - [Memory and state carried across sessions](https://feedbackbench.com/criteria/context.session_memory.md)
  - [Finding the right files in the codebase](https://feedbackbench.com/criteria/context.codebase_retrieval.md)
  - [Images, PDFs and file attachments as input](https://feedbackbench.com/criteria/context.attachments.md)
- [Doing the work](https://feedbackbench.com/criteria/work.md): How does it behave while it works?
  - [Can do the user's kind of task](https://feedbackbench.com/criteria/work.capability.md)
  - [Frontend and visual UI output](https://feedbackbench.com/criteria/work.frontend_ui.md)
  - [Diagnosing and fixing reported bugs](https://feedbackbench.com/criteria/work.bug_diagnosis.md)
  - [Breaks existing code or reintroduces bugs](https://feedbackbench.com/criteria/work.regressions_introduced.md)
  - [Does unrequested work or over-engineers](https://feedbackbench.com/criteria/work.scope_overreach.md)
  - [Spins, loops or gets stuck without progress](https://feedbackbench.com/criteria/work.stuck_loops.md)
  - [Stops mid-task or answers instead of acting](https://feedbackbench.com/criteria/work.premature_stop.md)
  - [Long unattended runs and goal/loop mode](https://feedbackbench.com/criteria/work.long_running_autonomy.md)
  - [Subagents, parallel agents and orchestrators](https://feedbackbench.com/criteria/work.multi_agent_orchestration.md)
  - [Games checks instead of fixing the problem](https://feedbackbench.com/criteria/work.reward_hacking.md)
  - [Risky or irreversible actions without confirmation](https://feedbackbench.com/criteria/work.destructive_actions.md)
  - [Git commits, branches and sync](https://feedbackbench.com/criteria/work.git_workflow.md)
  - [Computer use and browser control](https://feedbackbench.com/criteria/work.computer_browser_use.md)
  - [Safety filters block legitimate coding tasks](https://feedbackbench.com/criteria/work.safety_refusals.md)
  - [Tool approval prompts and autonomy modes](https://feedbackbench.com/criteria/work.permission_prompts.md)
  - [Plan-before-edit mode](https://feedbackbench.com/criteria/work.plan_mode.md)
  - [Length and clarity of replies, summaries and comments](https://feedbackbench.com/criteria/work.response_verbosity.md)
  - [Caves to or argues with the user's judgement](https://feedbackbench.com/criteria/work.sycophancy_pushback.md)
- [Checking and finishing](https://feedbackbench.com/criteria/checking.md): Can you trust that the work is done?
  - [Claims work is done or fixed when it is not](https://feedbackbench.com/criteria/verify.false_completion.md)
  - [Builds, tests or runs its own changes](https://feedbackbench.com/criteria/verify.self_testing.md)
  - [Agent-performed code review finds real issues](https://feedbackbench.com/criteria/verify.agent_code_review.md)
  - [Reviewing and approving the agent's changes](https://feedbackbench.com/criteria/verify.change_review_ui.md)
- [Interface and sessions](https://feedbackbench.com/criteria/interface.md): How do you operate and steer it?
  - [How the interface shows work, and what the user can configure](https://feedbackbench.com/criteria/ui.display_settings.md)
  - [Saving, switching, resuming and rewinding sessions](https://feedbackbench.com/criteria/ui.session_history.md)
  - [Stopping and steering a running agent](https://feedbackbench.com/criteria/ui.interrupt_steer.md)
  - [Mobile, remote-control and voice access](https://feedbackbench.com/criteria/surfaces.remote_mobile.md)
  - [Cloud and remote sandbox execution](https://feedbackbench.com/criteria/surfaces.cloud_sessions.md)
- [Reliability and speed](https://feedbackbench.com/criteria/reliability.md): Does it stay up and respond quickly?
  - [Outages, server errors and capacity or rate errors](https://feedbackbench.com/criteria/rel.service_errors.md)
  - [Latency, throughput and fast mode](https://feedbackbench.com/criteria/rel.response_speed.md)
  - [Client crashes, freezes and failed tool execution](https://feedbackbench.com/criteria/rel.client_failures.md)
  - [Updates break working setups](https://feedbackbench.com/criteria/rel.update_breakage.md)
- [Account and support](https://feedbackbench.com/criteria/account.md): How does the vendor treat your account?
  - [Support, refunds and issue handling](https://feedbackbench.com/criteria/account.support.md)
  - [Wrong charges, failed payments and plan provisioning](https://feedbackbench.com/criteria/account.billing_errors.md)
  - [Account bans and access restrictions](https://feedbackbench.com/criteria/account.bans_restrictions.md)
  - [Data retention, training use and deployment isolation](https://feedbackbench.com/criteria/account.data_privacy.md)

## Data

- [coding.json](https://feedbackbench.com/data/coding.json): the full build: meta (rules, parameters, weights, baselines, criteria) and every agent with every criterion, intervals and posts
- [rankings.csv](https://feedbackbench.com/data/rankings.csv): one row per agent
- [criteria.csv](https://feedbackbench.com/data/criteria.csv): one row per agent and area or criterion
- [requests.csv](https://feedbackbench.com/data/requests.csv): the top requests of every agent, area and criterion, counted in author-weeks

## Optional

- [Web page](https://feedbackbench.com/): the interactive version for people
- [Enterpret](https://www.enterpret.com): builds Feedback Bench; runs the same feedback analysis on a company's own customer feedback