DvalinCodeThe approvable coding agent for regulated teams

Any model · local-first · policy-bound · audit-ready — AI coding your security team can actually approve.

Org policy bounds the agent

A company — not the developer — constrains modes, shell commands, paths, tools, and models via dvalin.policy.json. A repo policy can only narrow the machine policy, never widen it.

Policy reference

🛡️

Tamper-evident audit trail

Every run emits a hash-chained JSONL log — every file read/written, every command, every approval. Verify the chain offline with `dvalincode report verify`.

Threat model

🏛️

Evidence, not claims

OpenSSF Scorecard, CodeQL, pinned Actions, ISO/IEC 42001 alignment docs, and an offline-verifiable Evidence Pack — maintained as reviewable project artifacts.

Evidence pack

🔑

Any model, no lock-in

DeepSeek, OpenAI, Claude via OpenRouter, Groq, Ollama, or any OpenAI-compatible endpoint. Switch with one click — run fully offline with local models.

💻

Local-first, zero-dep binary

One ~25 MB executable per platform. No Node, no Python, no Docker. Sessions, config, and audit logs stay in ~/.dvalincode on your machine.

🧰

Secure remediation built in

Scan locally or import SARIF from CodeQL, GitHub Code Scanning, or Semgrep — findings become isolated remediation worktrees with PR-ready reporting.

Workflow

Install in 60 seconds

Don't take the claims on trust — verify them on your own machine:

curl -fsSL https://raw.githubusercontent.com/arthurpanhku/dvalincode/main/scripts/install.sh | bash
dvalincode trust

trust prints this install's live security posture: the resolved org policy and its hash, per-boundary network enforcement (provider · shell · MCP), and the tamper-evident audit status — the exact evidence a security reviewer needs, straight from the tool itself.

Then let the agent work, and prove what it did after the fact:

dvalincode report verify    # re-derive the hash chain of the last run's audit log

Windows builds and manual downloads for every platform are on the releases page, with SHA256SUMS.txt and build provenance attestation for each archive.

One binary, three frontends

Run dvalincode bare for an interactive terminal agent with streaming output, inline approvals, and red/green diffs — or dvalincode serve to host the web GUI for browser and remote use. An experimental desktop app ships on a separate pre-release track. All three drive the same agent core.

DvalinCode web GUI

Built for teams that need a "yes" from security

DvalinCode is an approvable agent runtime, not just another coding agent. The product is the evidence a security, compliance, or platform team needs to safely allow AI coding in finance, healthcare, and other confidential codebases:

Controllable — an org policy bounds the blast radius.
Transparent — dvalincode trust makes the posture self-verifiable.
Auditable — the hash-chained log proves what every run did.

Start with the threat model to see the full attack surface — malicious AGENTS.md, poisoned MCP servers, prompt-injection escalation, egress, audit tampering — each mapped to the control that defends it and the honest residual gap.