# kilo-org-kilocode > Writes, edits, and debugs code across multiple files in any language using natural language instructions from VS Code, JetBrains, or the terminal. - repo: https://github.com/kilo-org/kilocode - commit: 24167bd703f56b8adc39cf94263cb85736963777 - api style: openai - runs: 8 (8 with a reward) - tasks tried: 7 - models: bedrock/us.anthropic.claude-haiku-4-5-20251001-v1:0 - domains: swe - languages: go, java, javascript, python, rust - capabilities: edits-files, runs-shell, runs-tests, browses-web, uses-git, multi-agent, long-horizon, reads-docs, calls-apis ## How it runs here Kilo CLI (packages/opencode, a fork of OpenCode; core services in packages/core) is a Bun/TypeScript coding agent. Its entrypoint is packages/opencode/src/index.ts (yargs CLI, script name `kilo`); the non-interactive one-shot mode is `kilo run --format json --auto --model provider/model "<prompt>"`, which boots the in-process HTTP server, creates a session rooted at the cwd, streams one JSON event per line, auto-approves permission requests, and exits. Model calls go through the Vercel AI SDK; providers come from JSON config (KILO_CONFIG_CONTENT holds inline config), and a custom provider with npm "@ai-sdk/openai-compatible" plus options.baseURL sends OpenAI chat-completions requests (streaming + tool calls) to the proxy, exactly the way the repo's own CLI subprocess tests do (test/lib/test-provider.ts, test/lib/cli-process.ts). The overlay installs Bun 1.3.14 under /opt/harness/bun, copies the repo to /opt/harness/src, runs `bun install --frozen-lockfile`, and writes a /opt/harness/bin/kilo shim (symlinked into /usr/local/bin and added via profile.d so `bash -lc` keeps it on PATH) that launches src/index.ts with the same bun flags as the repo's dev/test tooling, with the @opentui/solid preload resolved to an absolute path at build time and offline-safe KILO_DISABLE_* defaults exported so the CLI never fetches models.dev, plugins, or updates. ## Results by task | task | runs | last reward | best reward | last tests | |---|---|---|---|---| | [aider_polyglot/polyglot_go_palindrome-products](https://harnessreport.com/tasks/aider_polyglot/polyglot_go_palindrome-products.md) | 2 | 0 | 0 | | | [aider_polyglot/polyglot_python_bowling](https://harnessreport.com/tasks/aider_polyglot/polyglot_python_bowling.md) | 1 | 1 | 1 | 31 passed | | [algotune/algotune-vertex-cover](https://harnessreport.com/tasks/algotune/algotune-vertex-cover.md) | 1 | 1 | 1 | 2 failed, 1 passed | | [bigcodebench_hard_complete/bigcodebench_1057](https://harnessreport.com/tasks/bigcodebench_hard_complete/bigcodebench_1057.md) | 1 | 1 | 1 | 6 passed | | [bird-bench/card_games__372](https://harnessreport.com/tasks/bird-bench/card_games__372.md) | 1 | 1 | 1 | | | [humanevalfix/python-12](https://harnessreport.com/tasks/humanevalfix/python-12.md) | 1 | 1 | 1 | 1 passed | | [swebench-verified/psf__requests-5414](https://harnessreport.com/tasks/swebench-verified/psf__requests-5414.md) | 1 | 1 | 1 | 131 passed, 1 xfailed, 158 errors | ## Tests to run next _ranked by llm_ | task | why | |---|---| | [aider_polyglot/polyglot_rust_bowling](https://harnessreport.com/tasks/aider_polyglot/polyglot_rust_bowling.md) | Rust is in your language profile but untried; bowling scoring passed with 31 tests in Python, so this tests if you can generalize the same algorithmic solution across languages. | | [aider_polyglot/polyglot_javascript_triangle](https://harnessreport.com/tasks/aider_polyglot/polyglot_javascript_triangle.md) | JavaScript is in your language profile but untried; geometry classification is simpler than the Go palindrome task that failed, so this tests a cleaner baseline in a new language. | | [quixbugs/quixbugs-python-sieve](https://harnessreport.com/tasks/quixbugs/quixbugs-python-sieve.md) | New taskset focused on pure bug-fixing with one-line edits; tests your core capability on focused, minimal-blast-radius changes. | | [humanevalfix/python-21](https://harnessreport.com/tasks/humanevalfix/python-21.md) | You passed python-12 in this taskset; this is a different single-function fix to test consistency and surface any task-specific brittleness. | | [swebench-verified/sphinx-doc__sphinx-8595](https://harnessreport.com/tasks/swebench-verified/sphinx-doc__sphinx-8595.md) | New swebench task marked as <15 min fix; tests whether you can handle real codebase understanding on a different project than the requests fix that had test errors. | ## Runs | run | task | verifier says | calls | seconds | |---|---|---|---|---| | [20260925T233225-kilo-org-kil-d996c3e6b2fa-polyglot_go_palindrome-products](https://harnessreport.com/runs/20260925T233225-kilo-org-kil-d996c3e6b2fa-polyglot_go_palindrome-products.md) | aider_polyglot/polyglot_go_palindrome-products | reward 0 | 12 | 50 | | [20260925T230452-kilo-org-kil-ac8dba9d8b5d-bigcodebench_1057](https://harnessreport.com/runs/20260925T230452-kilo-org-kil-ac8dba9d8b5d-bigcodebench_1057.md) | bigcodebench_hard_complete/bigcodebench_1057 | reward 1 · 6 passed | 11 | 49 | | [20260925T224236-kilo-org-kil-b0bb044ff780-algotune-vertex-cover](https://harnessreport.com/runs/20260925T224236-kilo-org-kil-b0bb044ff780-algotune-vertex-cover.md) | algotune/algotune-vertex-cover | reward 1 · 2 failed, 1 passed · verifier exited 1 | 55 | 378 | | [20260924T231950-sw-kilocode-card_games__372](https://harnessreport.com/runs/20260924T231950-sw-kilocode-card_games__372.md) | bird-bench/card_games__372 | reward 1 | 4 | 16 | | [20260924T231500-sw-kilocode-python-12](https://harnessreport.com/runs/20260924T231500-sw-kilocode-python-12.md) | humanevalfix/python-12 | reward 1 · 1 passed | 7 | 23 | | [20260924T230539-sw-kilocode-psf__requests-5414](https://harnessreport.com/runs/20260924T230539-sw-kilocode-psf__requests-5414.md) | swebench-verified/psf__requests-5414 | reward 1 · 131 passed, 1 xfailed, 158 errors | 91 | 426 | | [20260924T225619-sw-kilocode-polyglot_go_palindrome-products](https://harnessreport.com/runs/20260924T225619-sw-kilocode-polyglot_go_palindrome-products.md) | aider_polyglot/polyglot_go_palindrome-products | reward 0 | 18 | 114 | | [20260923T134847-kilo-org-kil-polyglot_python_bowling](https://harnessreport.com/runs/20260923T134847-kilo-org-kil-polyglot_python_bowling.md) | aider_polyglot/polyglot_python_bowling | reward 1 · 31 passed | 54 | 317 | --- Harness Report runs agent harnesses from their GitHub repos on Harbor tasks and records every model call. Every page is also `.md` and `.json`; index: https://harnessreport.com/llms.txt · MCP: https://harnessreport.com/mcp