ctx

Find the cheapest AI coding setup that actually works on your repo.

CTX Fit analyzes your repository, tests promising AI coding configurations against real tasks in it, and produces the winning configuration as a reviewable change — in your working tree with --apply, or as a pull request with --pr. It picks the cheapest setup that reliably works — reliability is a requirement, not a tie-break — and if nothing beats what you already have, it says so.

The winner is chosen by a fixed rule, not a score: discard every candidate below the reliability floor, then minimize attributable cost, then break ties toward the simpler configuration. An LLM may explain a result; it never decides one.

Release scope (1.0.21). CTX Fit compares capability configurations within one coding-agent harness; it does not compare Codex, Claude Code, or other harnesses against one another. It recognizes and can run repository-native verification commands for Python, JavaScript/TypeScript, Go, Rust, and Make, and treats the selected test command as the verification authority. For an installable Python project, CTX Fit builds a campaign environment and installs it without network access; its build backend and dependencies must already be available without downloading them. In the other ecosystems, verification is supported only when the runtime is usable from the host PATH under an isolated home and the verification dependencies are already available in the repository. Final verification uses that isolated home and runs without network access, so a user's package caches are not a supported dependency source. This is evidence for normal development; it does not prove that deliberately hostile code cannot deceive its own test runner. Release qualification did not include a paid live-provider trial, so inspect ctx doctor and the dry run before authorizing spend.

pip install --upgrade claude-ctx
cd /path/to/my-project
ctx fit

Bare ctx fit is free, local and read-only: it runs no model, spends nothing, and issues no git commands at all. Example output, abridged from a real run against this repository:

Repository: /path/to/ctx
Languages:  python, javascript

Current AI coding setup
  Instructions:  AGENTS.md, CLAUDE.md
  Tool config:   .claude/settings.local.json
  Installed skills: 26

How this repository verifies itself
  test       python -m pytest -q
             from pyproject.toml [tool.pytest] (high confidence)
  typecheck  python -m mypy src
             from pyproject.toml [tool.mypy] (high confidence)
  lint       python -m ruff check .
             from pyproject.toml [tool.ruff] (high confidence)
  build      python -m build
             from pyproject.toml [build-system] (medium confidence)

AI agent readiness
  91/100
    Verification           30/30
    Instructions           20/20
    Environment             6/15
    CI enforcement         15/15
    Tool safety            10/10
    Context tractability   10/10

Highest-impact improvements
  1. Commit a dependency lockfile. (+9)
     no dependency lockfile is committed

This repository has the static evidence needed to plan an evaluation: it
declares deterministic tests. Whether those tests can execute is checked only
inside the campaign.

Requires Python 3.11 or newer. Add --json for machine-readable output, or --dry-run to see what a full evaluation would involve. --dry-run does read your history — it runs read-only git queries (log, show --name-only, ls-tree, rev-parse) to derive representative tasks — and writes nothing: not to the repository, not to the index, not to any ref.

Beyond the free profile, ctx fit --test --budget N evaluates candidate configurations against those tasks. Spending needs both flags: --test without --budget only plans. Run ctx doctor to see whether a real evaluation can run here. A real evaluation needs pip install "claude-ctx[harness]", Node.js with npx for the workspace-filesystem MCP, a matching provider credential, and Bubblewrap on Linux; the base install can profile, plan, and simulate. Without a matching provider credential, --test runs in simulation, which proves the pipeline but not your repository. With a credential but a missing live prerequisite, CTX refuses the run before trial setup. A simulated result is refused as evidence for --apply and --pr.

Ubuntu 24.04: enable Bubblewrap's packaged AppArmor profile

Ubuntu 24.04 restricts unprivileged user namespaces, and merely installing bwrap does not prove it can start the network-disabled namespace CTX uses for repository commands. Install and load Ubuntu's packaged, scoped bwrap-userns-restrict profile for /usr/bin/bwrap:

sudo apt update
sudo apt install bubblewrap apparmor-profiles apparmor-utils
if [ ! -e /etc/apparmor.d/bwrap-userns-restrict ]; then
  sudo install -m 0644 \
    /usr/share/apparmor/extra-profiles/bwrap-userns-restrict \
    /etc/apparmor.d/bwrap-userns-restrict
fi
sudo apparmor_parser -r /etc/apparmor.d/bwrap-userns-restrict
ctx doctor

Keep Ubuntu's global unprivileged-user-namespace restriction enabled; CTX uses the targeted Bubblewrap profile instead of weakening that system-wide security boundary. The profile is administrator-visible host policy for every /usr/bin/bwrap caller, not a CTX-private setting; the commands above preserve an existing local profile rather than overwriting it. ctx doctor proves this path with a bounded /bin/true probe in the same no-network namespace. It executes no repository code and calls no model. See Ubuntu's AppArmor user-namespace guidance and the packaged Bubblewrap profile.

--apply and --pr write different things

ctx fit --apply writes the winning configuration into your working tree, on whatever branch you are standing on. It prints every proposed change first and stops there unless you pass --yes. The write itself runs no git command: nothing is staged, committed, or pushed. Getting to it does run git — --apply is refused without evidence from ctx fit --test --budget N, and deriving the tasks for that evaluation uses the same read-only queries --dry-run uses.

Each proposed change names the file and whether CTX Fit is creating or modifying it. Today every plan contains exactly one CTX-owned artifact, .ctx/fit-configuration.json. The sidecar records the pinned model plus the exact instruction and capability bytes that were evaluated, with their hashes; ordinary ctx run invocations validate and activate that configuration.

CTX Fit does not rewrite AGENTS.md, CLAUDE.md, or other user-authored instruction files. Their evaluated bytes are embedded in the sidecar instead. If an existing untracked sidecar matters to you, save a copy before confirming the write; version-control restore commands cannot recover an untracked file.

ctx fit --pr writes to a remote. It creates a branch, commits the winning configuration, pushes it to origin, and opens a pull request through the GitHub CLI. Before running anything it prints the pull-request body, the files it will write, and the exact command sequence:

git checkout -b ctx-fit/<timestamp>
git add -- <paths>
git commit -m "<pull request title>"
git push --set-upstream origin ctx-fit/<timestamp>
gh pr create --title "<pull request title>" --body-file -

Without --yes it stops there and changes nothing. With --yes it writes those files into the working tree and then runs those five commands, in that order and no others. Before any of them runs, the gate described below runs read-only probes — git rev-parse, git status, git remote get-url, and gh auth status — which is what lets every refusal leave the repo