Integrations

Multiple operator surfaces, one receipt format: CLI, VS Code, CI/CD, and AI assistants via MCP.

Choose the surface that fits your workflow

Every AIIR surface generates the same receipt format. Choose the operator surface that fits your workflow.

💻 CLI

The zero-dependency CLI path for first receipts, history backfills, and direct verification.

🧩 VS Code

The local-first operator surface for one repo: record the current commit, inspect the receipt, and carry editor context directly into the proof.

⚙️ CI/CD

The repeatable pipeline surface for team-wide coverage, policy gates, and optional Sigstore-signed release evidence.

🤖 AI assistants via MCP

The assistant surface for Claude, Copilot, Cursor, and other MCP clients that call the same local AIIR installation.

Ecosystem adapters

These operator surfaces are the core product. Around them, AIIR can also fit into existing attestation, provenance, and policy ecosystems as one more receipt-bearing surface rather than a replacement layer.

Public implementation notes These notes are intentionally narrow and T0-safe. They show how AIIR can be carried through external systems without claiming official support or standards status before that work exists.

Witness attestor

Emit an AIIR receipt as one more artifact in a Witness-style pipeline.

Read the note →

GUAC ingestor

Make AIIR receipts queryable as source provenance metadata in a graph.

Read the note →

In-toto policy input

Carry AIIR in an in-toto Statement and use receipt verification as one narrow policy signal in engines such as AMPEL.

Read the note →

Open ecosystem adapters →

CI/CD Platforms

AIIR ships native integrations for GitHub Actions and GitLab CI/CD, plus ready-to-use YAML templates for Azure Pipelines, Bitbucket, CircleCI, and Jenkins. Receipts are generated at build time and stored as pipeline artifacts — no external service required.

⚙️ GitHub Actions

Native composite action with SHA-pinned dependencies, OIDC Sigstore signing, and artifact attestation support.

invariant-systems-ai/aiir@v1 (SHA-pin recommended for production)

🦊 GitLab CI/CD

Published CI/CD Catalog component. One-line include: in your pipeline. Duo Chat integration via MCP. AIIR is a GitLab Technology Partner.

CI/CD Catalog · Component · GitLab Partner

🔷 Azure Pipelines

Pipeline template with receipt generation, verification gate, and artifact publishing. Drop into any Azure DevOps project.

YAML template

🪣 Bitbucket Pipelines

Pipe-compatible step definition. Generates receipts and attaches them as build artifacts in Bitbucket Cloud.

YAML template

🔵 CircleCI

Orb-compatible job definition with receipt generation and artifact storage. Works with CircleCI Cloud and Server.

YAML template

🏗️ Jenkins

Declarative pipeline step. Install via pip in your agent, call aiir as a shell step.

Jenkinsfile template

All platforms produce the same receipt format. The AIIR receipt format is platform-independent. A receipt generated by GitHub Actions uses the same schema, verification logic, and content-addressing as one generated by Jenkins or locally. The generator field records which integration produced it. Verify anywhere: aiir --verify receipt.json

VS Code Extension

The AIIR extension for VS Code brings receipt verification, generation, health checks, and posture management directly into the editor. It runs local-only by default — no network calls unless you opt in.

📜 Receipt Explorer

Browse receipts grouped by commit with inline verification badges. Expand any receipt to inspect overview, signals, files, artifacts, and provenance.

🤖 Editor Context Capture

Observes active AI coding extensions (Copilot, Cline, Cursor, Tabnine, and 10+ others) and can pass that local context to the CLI when you generate a receipt.

✅ Inline Verification

CodeLens and tree-view actions for verifying receipt integrity, CBOR sidecar hashes, and Sigstore signatures — all without leaving the editor.

🛡️ Security Posture

Read-only posture view for local-only mode, workspace isolation, allowlists, and CLI/Sigstore coverage. Deployment presets for team rollouts.

🔄 Auto-Receipting

Managed post-commit hook installs with one command. Observed local AI tool context can be attached automatically on every commit.

🏠 Home View

Human-first landing page showing trust status, current HEAD coverage, next best action, and the fastest path to value for new users.

Install from the Marketplace or side-load the VSIX from the repo. The extension requires the AIIR CLI for receipt generation: pip install aiir.

AI Assistants via MCP

AIIR ships an MCP server that works with major AI coding assistants. This is an assistant-facing control surface on top of the same local AIIR install, not a separate provenance system.

GitHub Copilot

VS Code MCP integration. Supports org-level MCP registry for enterprise rollout.

Claude

Claude Desktop and Claude Code. One-click MCP server configuration.

Cursor

Native MCP support. Add to your Cursor MCP config and use in any project.

Windsurf

MCP server integration. Works across Windsurf IDE and browser clients.

Continue

Open-source AI assistant. MCP server support via the VS Code Marketplace extension.

Cline

MCP Marketplace compatible. Works with Cline's autonomous coding workflow.

MCP Server Tools

Seven tools exposed over the standard MCP protocol:

Tool	Description
`aiir_receipt`	Generate a receipt for any commit or range
`aiir_verify`	Verify receipt integrity (content hash + receipt ID)
`aiir_stats`	Ledger statistics — receipt count, AI %, date range
`aiir_explain`	Human-readable verification explanation
`aiir_policy_check`	Evaluate ledger against policy constraints
`aiir_verify_release`	Produce a Verification Summary Attestation (VSA) for CI gates
`aiir_gitlab_summary`	GitLab Duo Chat / MR comment summary with optional SAST data

Install once, use everywhere. pip install aiir installs both the CLI and the MCP server. Run aiir-mcp-server --stdio or configure your assistant with the snippet from Docs.

Evidence depth by surface

Different surfaces produce different evidence strengths. The important point is that they still verify the same way.

Surface	Default evidence	Stronger path
CLI	Declared commit evidence	Run in CI later or attach signed release evidence
VS Code extension	Active editor context	Deterministic editor provenance when AIIR directly observed edits
CI/CD	Repeatable build-time receipt generation	Sigstore-signed CI evidence
AI assistants via MCP	The same local CLI evidence, invoked through an assistant	Pair with VS Code or CI when you need richer attestation

See Docs for the full evidence-level model and Verify for verification tiers.

Declared Signals and Tool Context

AIIR understands declared AI signals from 16+ coding assistants and related automation markers. It reads commit metadata, trailers, branch names, and authorship patterns, and local integrations can add active editor context when that is available.

Tool	Examples of context AIIR can normalize
GitHub Copilot	Commit trailers, author patterns, branch names
ChatGPT / OpenAI	Commit messages, author metadata
Claude / Anthropic	Author patterns, commit metadata, trailers
Cursor	Author patterns, cursor-specific trailers
GitLab Duo	Duo Code Suggestions, Duo Chat, co-authored-by trailers
Amazon Q / CodeWhisperer	Commit trailers, author metadata
Google Gemini	Author patterns, commit metadata, code assist signals
Devin	Author patterns, co-authored-by trailers, commit metadata
Tabnine	Commit trailers, authorship signals
Codeium / Windsurf	Author metadata, trailers
Sourcegraph Cody	Author patterns, commit metadata
Aider	Commit message prefix patterns
Replit AI	Author patterns, commit metadata
JetBrains AI	Author patterns, commit trailers
Supermaven	Author patterns, commit metadata
bolt.new / Lovable	Author patterns, commit metadata

Signal normalization is best-effort and documented honestly. AIIR does not claim to prove hidden AI usage. See Docs and Trust for the exact boundaries.

Attestation & Verification

AIIR receipts are designed to feed into the broader software supply chain attestation ecosystem.

in-toto Statements

Verification Summary Attestations (VSA) wrap AIIR results as in-toto v1 Statements. Compatible with SLSA and any in-toto verifier.

Sigstore Signing

Keyless signing via Fulcio + Rekor transparency log. Non-repudiation without key management. Optional — works without signing too.

GitHub Attestations

AIIR SBOMs and attestations integrate with GitHub's gh attestation verify flow and predicate-type filtering.

Policy Engines

AIIR verification results are structured JSON — ready for OPA/Rego policies, Kyverno admission controllers, or any policy-as-code system.

How it fits together

AIIR is both an emitter and a verifier. In a typical deployment:

Developer commits with declared AI context →
CI pipeline runs aiir → receipt generated →
Receipt stored as artifact →
aiir --verify-release --policy balanced →
VSA attestation produced →
Policy engine evaluates VSA → deploy gate decision

Every step uses the same open format (AIIR Commit Receipt v2). Every step can be verified independently. No vendor lock-in at any point.

Get started

pip install aiir          # CLI + MCP server, zero dependencies on the CLI path
aiir --pretty             # Receipt your latest commit
aiir --verify receipt.json  # Verify any receipt

Read the docs → View on GitHub → GitLab Catalog →