DevTime

DevTime verifies what your repository can actually prove.

A file named stripe/webhook.ts looks like proof that a repo handles Stripe webhooks. It might be a handler that only returns 404. DevTime checks statements about a repository against its implementation, tests, configuration, and recorded decisions, then reports what is supported, what is contradicted, what is missing, and what went stale.

No cloud. No telemetry. No code execution. No AI required.

Prefer video? Watch the 2-minute demo: DevTime scans a repo locally, explains concepts from evidence, surfaces uncertainty, catches a risky diff, and shows how a corroborated decision improves understanding.

Try DevTime in 60 seconds

pipx install devtime-ei
dtc demo init
cd devtime-demo-saas
dtc init
dtc scan
dtc verify

On the demo repo that ends with billing webhook signature verification SUPPORTED, JWT authentication SUPPORTED, and route test association WEAK, naming the routes with no test importing them. Point it at your own repository and the answers change:

cd your-repo
dtc init && dtc scan && dtc verify

The PyPI distribution is devtime-ei. The Python package remains devtime, and the CLI command remains dtc. dtc demo init copies a small static example repo into ./devtime-demo-saas so you can try DevTime without cloning this repository.

From source

git clone https://github.com/Shakargy/devtime.git
cd devtime
python -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"
cd examples/demo-saas
dtc init
dtc scan
dtc concepts
dtc explain "Billing Webhooks"

On Windows PowerShell:

git clone https://github.com/Shakargy/devtime.git
cd devtime
python -m venv .venv
.venv\Scripts\Activate.ps1
pip install -e ".[dev]"
cd examples/demo-saas
dtc init
dtc scan
dtc concepts
dtc explain "Billing Webhooks"

You should see Billing Webhooks explained from evidence, including supported claims, file references, uncertainty, Understanding Score, and Understanding Debt.

To test risk review, make a local change first, then run:

dtc risk --diff

A full, copy-pasteable walkthrough (including the risk-diff and corroborated-decision steps) is in DEMO_SCRIPT.md.

Verify claims (experimental)

A claim is a statement about the repository. Verification answers it with a status and receipts, never with confidence the evidence cannot back.

Status: CONTRADICTED

Contradictions:
  - The billing webhook endpoint cannot verify signatures
    because it is a disabled stub.
      claimed:  pages/api/stripe/webhook.ts is named and routed
                as a billing webhook endpoint.
      observed: The handler's only behavior is a 404/501 response.

• SUPPORTED - required behavior evidence exists in the current scan
• WEAK - the surface exists, but the proving evidence is missing
• CONTRADICTED - credible evidence conflicts with the claim, both sides shown
• UNKNOWN - the surface exists but coverage cannot responsibly decide
• NOT_APPLICABLE - the repository has no surface this claim is about

Four built-in claims ship: route test association, admin authorization, billing webhook signatures, and JWT authentication. dtc verify leads with what it can actually verify in your repository, and when nothing applies it says what would make a claim verifiable instead of dead-ending.

Truth and freshness are separate: when a file behind a verified claim changes, the claim goes STALE and names the file.

To review a change, compare claims at two commits:

dtc review --base origin/main

It reports what the change did to each claim (a regression, a claim that newly applies, evidence that changed under a claim that held) from isolated snapshots of both commits, without touching your working tree. It runs in GitHub Actions as an advisory job summary.

See VERIFICATION.md for the full model and its limits.

Why this exists

Git records what changed, but it does not preserve the reasoning behind those changes. When you return to a repository - or review one you did not write - you often have to reconstruct why a behavior exists, what evidence supports it, and what is still uncertain.

DevTime builds evidence-backed repository memory: a local layer that helps a codebase explain itself from code, tests, configs, routes, and recorded decisions. It shows what the repository can support with evidence - and, just as importantly, what it cannot support yet.

Who it is for

DevTime is for people who need to understand a repository from evidence rather than memory.

It is especially useful if you:

• are onboarding to an unfamiliar codebase and need to understand how a feature is implemented;
• are reviewing a pull request and want to see what evidence supports a behavior;
• are returning to a project after weeks or months and cannot remember why something exists;
• maintain a long-lived project where design decisions are easily lost;
• want repository understanding to be backed by code and recorded decisions instead of generated summaries.

Questions DevTime helps answer include:

• Can this repository actually prove the thing its file names imply?
• Where is authentication actually implemented?
• What files prove that Billing Webhooks exist?
• What is still uncertain?
• Did this diff touch a risky concept?
• Is there a decision explaining this behavior?

What DevTime does

• Verifies claims about a repository and reports status, evidence, and both sides of any contradiction.
• Tracks freshness, so a verified claim goes stale when the evidence behind it changes.
• Detects concepts from routes, tests, configs, dependencies, and docs.
• Explains from evidence by linking claims to files and signals.
• Surfaces uncertainty when evidence is missing or weak.
• Scores understanding with an Understanding Score and Understanding Debt label.
• Reviews narrow risky diffs with advisory findings, including which verified claims a diff destabilizes.
• Records decisions locally so rationale can reduce uncertainty when corroborated by code.

Supported concepts

Underneath verification is a scanner that builds local, evidence-backed memory:

DevTime detects six supported concept families. It does not discover arbitrary domain concepts yet:

• Authentication
• Billing Webhooks
• Background Jobs
• Data Export
• Admin Permissions
• File Uploads

Anything outside these six is out of scope for now. See LIMITATIONS.md.

What DevTime does not do

• It does not execute your code.
• It does not send code or data over the network.
• It does not require or call an AI model.
• It does not guarantee correctness or safe changes.
• It does not replace code review or architecture decisions.
• It is not a documentation generator, a static analyzer, an observability tool, a productivity tracker, or an AI coding agent.

Trust model

• DevTime stores local repository memory in .devtime/ (a local SQLite database).
• No network access during a scan.
• No code execution during a scan.
• Ignored directories are pruned before scanning; ignored files and secrets must never become evidence or claims.
• Every claim must link to evidence - no claim without evidence.
• Weak evidence produces uncertainty, not confidence.
• Usage is not decision: that a dependency is used does not mean someone decided why.
• Risk review is advisory by default - it does not block PRs.

Commands

(Also available: dtc evidence, dtc debt, dtc status, dtc doctor --privacy, dtc export, dtc reset, dtc mcp start.)

Requires Python >= 3.11 and git. See QUICKSTART.md for a step-by-step first run and troubleshooting.

Use with coding agents (MCP)

Your coding agent starts every session amnesiac about your repository and then guesses, confidently. DevTime gives it memory it can trust: a local, read-only MCP server that answers only with claims the repository can prove, plus explicit uncertainty for what it cannot.

Install with MCP support and scan your repo:

pipx install "devtime-ei[mcp]"
cd your-repo
dtc init
dtc scan

Add DevTime to Claude Code:

claude mcp add devtime -- dtc mcp start

Or in any MCP client that reads .mcp.json:

{
  "mcpServers": {
    "devtime": {
      "command": "dtc",
      "args": ["mcp", "start"]
    }
  }
}

The agent gets four read-only tools: list_concepts, explain_concept, get_context_pack (governed context with do-not-change-without-review paths, tests to run, and agent guidance), and verify_claim (claim status, contradictions, and missing evidence, computed fresh and never persisted). Local stdio only - no network listener, no write tools, no source code returned, only evidence file paths.