lanegate 1.0.3

Git-native agentic delivery engine — queue, parallel agents, quality gates, staged deployment

LaneGate

Parallel lanes. Protected gates. Every agent gets a lane. Nothing ships without a gate.

LaneGate helps coding agents (Claude, Codex, Ollama, aider, etc.) work on the same repo with less manual juggling. Tickets live as local Markdown files, agents claim them, worktrees keep changes separate, and file-level locks reduce overlapping edits. No SaaS, no subscriptions, no external project state.

LaneGate is not a replacement for coding agents, IDEs, or spec-driven planning tools. Those tools do the planning and implementation work. LaneGate records scoped work as repo-local state and coordinates the execution loop around it.

Security note. V1 provides git-level isolation and diff inspection. It does not sandbox agents at the OS level. Agents run as host processes. See Security Status and Known Limitations before running it on repositories you care about.

---

Demo

---

When LaneGate Helps

LaneGate is useful when you already use coding agents and want a small local workflow around them:

• keep agent work as local Markdown tickets
• run each task in its own git worktree
• reduce edit collisions with explicit file-level touches
• block out-of-scope or sensitive-file changes from auto-merge
• run configured safeguards before completion or merge
• pause for human review before anything lands on main

It is intentionally smaller than a company issue tracker and less ambitious than an IDE. Use those for planning, discussion, and implementation. Use LaneGate when you want repo-local execution state around agent runs.

By default, LaneGate commits ticket specs and review verdicts to git as they change. Tickets are a git-native artifact, not local-only state. Executor transcripts, logs, cooldowns, locks, and worktrees stay local under .lanegate/. Set commit_status_changes: false to opt back into fully zero-footprint local state.

MCP is built in. Run lanegate mcp to expose LaneGate commands as native tools for any MCP-compatible agent (Claude, Cursor, etc.), with no shell commands required.

LaneGate's loopback Python API (lanegate api) is built and running today, serving board, ticket, diff, and orchestration-run state as JSON/SSE. The first local UI is still planned as a small add-on launched with lanegate ui: a bundled TypeScript frontend over that API for board scanning, ticket detail, blocked/review triage, diffs, orchestration logs, and read-only settings preview. The CLI remains the complete fallback for advanced or custom workflows. The UI is not a SaaS service and does not move project state out of your checkout.

---

Concepts in 30 seconds

.lanegate/tickets/TICK-007.md   ← source of truth (YAML frontmatter)
        │
        ▼
lanegate next         ← agent asks: what should I work on?
lanegate start TICK-007  ← agent claims it; worktree + branch created
  ... agent works in .lanegate/worktrees/tick-007/ ...
lanegate complete TICK-007
lanegate review TICK-007
lanegate merge TICK-007   ← branch → main, worktree removed
lanegate promote production  ← staged deploy with guard/pre/post hooks

Main command groups:

Group	Commands
Scoping: record work as tickets and ask an executor to infer touches / close_criteria	create, analyze
Execution: who works on what without colliding	board, next, start, complete, review, merge
Delivery: how merged code reaches environments	promote, pipeline-status, flag

---

Security Status

LaneGate is experimental software for coordinating coding agents on a local repo. Before letting it run without review, read this section.

What LaneGate does to limit agent scope:

• Runs each agent in a dedicated git worktree, one per ticket
• Enforces a touches list: files committed outside the declared scope route to needs_review
• Hard-blocks writes to CI/CD configs, dependency manifests (Python, JS/Node, Rust, Go, Java/Gradle, Ruby, PHP, .NET), and credential-shaped files (.env, *.pem, secrets.*)
• Scans ticket text for prompt injection signals before dispatch
• Optionally runs gitleaks, semgrep/bandit, pip-audit, npm-audit, composer-audit, and bundler-audit on agent-produced diffs
• Runs configured safeguards such as tests or scripts before complete and merge

What LaneGate does NOT do:

• MCP is not a sandbox. lanegate mcp exposes ticket lifecycle commands as native tools, and an agent with access to those tools can claim tickets, create branches, and trigger merges. The MCP server authenticates nothing beyond what the MCP client config enforces.
• Host executors inherit host permissions. Agents (Claude Code, aider, Codex) run as the invoking OS user with full filesystem and network access. There is no bwrap, seccomp, or container wrapping.
• The file-based lock is single-machine, single-checkout. Two separate clones on two machines can both claim the same ticket.
• File-level locks are not semantic dependency analysis. Two tickets can touch different files and still break each other through shared APIs, imports, types, schemas, or runtime behavior. Treat touches as a practical coordination boundary, not a proof of correctness.

Recommendation: use --human-review final when running lanegate orchestrate on any repository with production code. This stops after implementation and requires lanegate merge <id> to be run manually after you inspect the diff. To make this the safe default without depending on every invocation remembering the flag, set default_human_review: final (or per_ticket) in .lanegate.yml. An explicit --human-review flag still overrides it. Note that this is unrelated to a ticket's autonomy field, which only governs auto-fix-retry behavior, not the merge gate.

For the full threat model, executor permissions, and safe usage notes, see SECURITY.md and docs/security-model.md.

---

Known Limitations

• Lock scope is single-machine, single-checkout. The file-based concurrency lock at .lanegate/orchestrator.lock prevents two lanegate orchestrate runs on the same checkout from racing, but does not coordinate across separate clones or machines.
• No OS-level sandbox in V1. Agents run as child processes with full user permissions (no bwrap, seccomp, or container wrapping), regardless of which Claude Code permission mode is configured (see the next bullet for that separate, application-level choice). LaneGate inspects the git diff after the agent exits. It cannot observe what the agent read or sent over the network during execution.
• Executor permissions come from the executor runtime, not from LaneGate. lanegate init configures Claude Code with a scoped --allowedTools set by default so the agent can run headless without interactive prompts, while tools outside that list stay gated. You can instead configure flags: ["--dangerously-skip-permissions"], which disables Claude Code's per-action approval prompts entirely. That's a valid choice for setups like a sandboxed CI runner, but it means the agent acts on anything without confirmation. See Security Status for the full set of options.

---

Platform support

Platform	Status
Linux	Supported, primary development platform
macOS	Supported, CI-verified on every push
Windows	Core functionality works, but projects using .sh executor scripts require WSL

---

Install

pip install lanegate
lanegate init    # scaffold .lanegate/ + .lanegate.yml in your repo

lgt is installed alongside lanegate as a short alias for the same command (e.g. lgt board, lgt next).

If your project has non-Python source (JS/TS, Go, Rust, Java, Ruby, C/C++), install the treesitter extra too:

pip install "lanegate[treesitter]"

Ticket analysis matches candidate files by parsing real symbols — via stdlib ast for Python, and tree-sitter for everything else. Without the extra, non-Python files fall back to a plain ripgrep text search instead of real symbol matching, which is less precise and pushes more exploration (and token cost) onto the analyzing agent. Python-only projects don't need it — stdlib ast covers .py files either way.

LaneGate is a standalone CLI, not a library you import into a project. pipx is the recommended way to install it, since it isolates the tool's own dependencies from whatever Python environment your project uses:

pipx install lanegate
lanegate init

(pipx install "lanegate[treesitter]" for the non-Python symbol matching noted above.)

To run from source:

git clone https://github.com/sudheerdvn/lanegate
cd lanegate
pip install -e ".[dev]"

---

Quick start

For real repositories, configure safeguards before you let an executor run:

safeguards:
  pre_complete:
    - pytest
  pre_merge:
    - pytest
  post_merge:
    - pytest

# 1. Create a ticket file
cat > .lanegate/tickets/TICK-001.md << 'EOF'
---
id: TICK-001
title: Add rate limiting to the upload endpoint
status: open
priority: 1
touches:
  - src/upload.py
  - src/middleware.py
parallel_safe: true
autonomy: supervised
close_criteria: POST /upload returns 429 after 100 requests/min per user
---

## Background
A single client can saturate the worker pool with rapid uploads.
EOF

# 2. See the board
lanegate board

# 3. Agent claims and works
lanegate start TICK-001
# ... implement in .lanegate/worktrees/tick-001/ ...
lanegate complete TICK-001
lanegate review TICK-001
lanegate merge TICK-001
lanegate validate TICK-001
lanegate done TICK-001

# 4. Deploy
lanegate promote production

---

Commands

Board & planning

lanegate board                    # ticket board + pipeline snapshot
lanegate next                     # recommend next ticket(s) + non-overlapping batch
lanegate pipeline-status          # commits pending at each environment stage

# Machine-readable output (any board/next/pipeline-status/flag list command):
lanegate --json board
lanegate --json next
lanegate --json pipeline-status
lanegate --json flag list

Ticket lifecycle

# Draft → open
lanegate analyze TICK-007         # fill touches + close_criteria, then open the ticket
lanegate open TICK-007            # flip draft → open without re-running analysis (requires touches already set)

# Execution
lanegate start TICK-007           # claim; creates branch + worktree
lanegate complete TICK-007        # code done → code_complete
lanegate review TICK-007          # submit for review → in_review
lanegate merge TICK-007           # merge to main, delete worktree → merged
lanegate validate TICK-007        # run configured post-merge checks → validated
lanegate done TICK-007            # close ticket → done

Deployment

lanegate promote staging          # guard → pre-promote → sync → post-promote
lanegate promote production
lanegate pipeline-status          # see what's pending where

Feature flags

lanegate flag list                        # show all flags (global)
lanegate flag list --env staging          # show flags for 'staging' environment
lanegate flag enable  new_checkout_flow --env staging
lanegate flag disable new_checkout_flow --env production

MCP server

lanegate mcp    # start stdio MCP server; attach any MCP-compatible client

Exposed tools include board, next_ticket, pipeline_status, flag_list, flag_set, repo_status, recent_logs, continuation_context, start, orchestrate, complete, review, merge, promote, hibernate, needs_review, fail, reopen, validate, done, and stats. Agent-facing action and log tools are bounded: lifecycle output is byte-capped, log excerpts are line- and byte-capped, and continuation state is reconstructed from durable repo data instead of chat context.

Local API / UI preview

lanegate api is built: a loopback-only (127.0.0.1) JSON/SSE server over board, ticket, diff, and orchestration-run state. lanegate ui (a bundled UI served by that API) is still planned:

lanegate api    # built — start the loopback JSON/SSE API for local clients
lanegate ui     # planned — start the API, serve the bundled UI, open a browser

Current lanegate api surface:

• GET /api/board
• GET /api/tickets
• GET /api/tickets/{id}
• GET /api/blocked
• GET /api/diff/{id}
• GET /api/status
• GET /api/config
• GET /api/pools
• PUT /api/pools/{name}/executors (reorders a pool's executor list)
• GET /api/runs
• GET /api/runs/current
• GET /api/runs/current/logs
• GET /api/runs/current/logs/stream
• POST /api/orchestrate/start
• POST /api/orchestrate/stop

Per-ticket lifecycle mutation endpoints (start/complete/review/merge) are still design-only. See docs/v2-interface-boundaries.md.

lanegate ui will use this API for common board, ticket, diff, review, and orchestration operations once built. Advanced or executor-specific workflows stay available through the CLI instead of being forced into the browser.

Terminal UI (TUI)

A Go TUI is also available via lanegate tui, rendering board, ticket detail, blocked queue, diff, orchestration-run, and settings screens from a fixture or the local API. It is mostly read-only, with one exception: the settings screen can reorder a pool's executor list and persist it via PUT /api/pools/{name}/executors. See docs/v2-interface-boundaries.md for its current scope and limits.

lanegate tui is a separate Go binary, not part of the lanegate Python package. A plain pip install does not give you lanegate-tui. To use it:

• Build from source (this repo ships the Go module at tui/): go build -o lanegate-tui ./tui/cmd/lanegate-tui, then either put lanegate-tui on your PATH or set LANEGATE_TUI_BIN=/path/to/lanegate-tui.
• Run without building, if you have the Go toolchain and a checkout of this repo: lanegate tui falls back to go run ./cmd/lanegate-tui automatically when no lanegate-tui binary is found on PATH or via LANEGATE_TUI_BIN.

If neither a binary nor the Go source is available, lanegate tui exits with an error telling you to set LANEGATE_TUI_BIN. See Troubleshooting.

Monitoring & auto-resume

Three detached background daemons cover runs you're not watching live, each with its own PID/log file under .lanegate/ and a --status/--stop pair:

lanegate watch                  # poll PR review decisions, auto-merge on approval
lanegate resume-watch           # wait out a rate limit on backoff, auto-retry `lanegate orchestrate`
lanegate resume-watch --history # what happened: hibernated -> retrying -> resumed/gave_up, with timestamps
lanegate notify-watch --test    # send a test phone push (ntfy.sh), verify setup
lanegate notify-watch           # push a phone alert when orchestrate looks stuck (dead process, stale
                              # heartbeat, or halted with tickets waiting)

See Rate limits and auto-resume and Phone alerts for stuck runs in the config reference for configuration and setup, including running notify-watch under systemd so it survives reboots.

Utilities

lanegate init             # scaffold .lanegate/ + .lanegate.yml
lanegate install-agent-tools # install Claude commands plus Codex/generic MCP snippets
lanegate install-commands # compatibility alias: copy Claude slash commands only
lanegate gh-sync          # mirror tickets to GitHub Issues (manual visibility sync; read-only view, not bidirectional)
lanegate gh-sync --dry-run  # preview what would be created or updated

---

Configuration

.lanegate.yml in your repo root, with walk-up discovery so it works from any subdirectory:

Executors

LaneGate dispatches tickets to whichever executor you configure. Each executor CLI has its own flag for non-interactive operation. Pass those flags via executors.<name>.flags so the orchestrator can run without waiting for prompts:

executor: claude   # default executor for orchestrate

executors:
  claude:
    flags: ["--allowedTools", "Bash,Edit,Write,Read,Glob,Grep"]  # scoped headless default; see Known Limitations
  aider:
    flags: ["--yes-always"]                    # auto-confirm all prompts; skip editor
  codex:
    flags: ["--approval-policy=never"]         # no per-action approval prompt
  ollama:
    models:
      implement: llama3.1
      review: qwen2.5-coder

Per-executor bin overrides the binary name for a supported executor. flags are prepended before the prompt.

For V1.5 multi-executor routing, set per-step executors:

executor: claude
executor_steps:
  implement: codex
  review: claude

When implement and review resolve to the same executor, LaneGate uses combined mode. When they differ, LaneGate implements first and then runs the resolved review route. A ticket-level executor: overrides implementation only. reviewer: controls review routing, including reviewer: human. Multi-executor routing is not OS/container sandboxing and does not isolate tools from the checkout.

For a batch-level human gate, run lanegate orchestrate --human-review final. For per-ticket human verdicts, set reviewer: human and run lanegate orchestrate --human-review per_ticket. LaneGate will wait for lanegate review <ticket> --verdict approved or a changes-requested verdict before merge.

Use safeguards to run deterministic checks before a ticket is marked complete, merged, or closed after merge:

safeguards:
  pre_complete:
    - pytest
  pre_merge:
    - pytest
    - scripts/pre-deploy-check.sh
  post_merge:
    - pytest

Supported guard commands include pytest, npm test, cargo test, go test, make ..., and executable scripts. A failing guard blocks the transition. post_merge runs from the merged control checkout during lanegate validate, and when it is configured, lanegate done requires validation first.

pre_merge guards also re-run automatically a second time, right after the git merge lands, against the merged main checkout itself, not just the ticket's isolated worktree. This catches the case where two independently-approved tickets each pass their own worktree's tests but break something once combined. If that second run fails, lanegate merge resets main back to its pre-merge commit and routes the ticket to needs_review instead of leaving a broken commit on main. See config-reference.md for details. This does not replace running the full suite on main yourself after merging a whole backlog in one sitting, it only re-checks after each individual merge.

For fully local/offline runs, use executor: aider with a local Ollama model as aider's backend. This is the validated path, since Aider has a real read/edit/commit loop:

executor: aider
models:
  analyze: llama3.1        # any executor works for analyze — text-only, no file edits
  implement: qwen2.5-coder # passed to aider as --model ollama/qwen2.5-coder
  review: qwen2.5-coder

Aider talks to Ollama's own local API. Install and pull those models with Ollama first (locally, or on whatever host aider's model route points at). See executor-capabilities.md for context_window_tokens and other local-model-specific settings.

executor: ollama (type: ollama) is a separate, more limited option: LaneGate posts directly to Ollama's REST API ({base_url}/api/generate}) and gets back a single text completion. Ollama itself has no file-editing or commit capability, so this cannot complete an implement/review step (it will produce zero commits and fail). It's only useful for text-only steps like analyze. See executor-capabilities.md for the full caveat.

For a rented/remote GPU box (e.g. a cloud instance running Ollama), tunnel its port instead of exposing it, since Ollama's API is unauthenticated:

ssh -N -L 11435:localhost:11434 <user>@<remote-host>
export OLLAMA_API_BASE=http://localhost:11435   # aider reads this directly; LaneGate has no separate field for it
lanegate orchestrate

Use verification.groups to tell the implement/review prompts when a UI check is expected and where to find the running app. LaneGate never runs a browser itself. The executor needs its own tooling (a Playwright MCP server, an in-session browser subagent) wired up separately. groups is a list because one repo can have more than one UI area, each with its own dev server and URL: separate frontend apps in a monorepo, or e.g. an AEM ui.frontend clientlib served by webpack alongside the actual AEM author instance:

verification:
  groups:
    - patterns: ["apps/web/**", "**/*.tsx"]
      dev_server: "npm run dev"
      url: "http://localhost:3000"
    - patterns: ["apps/admin/**"]
      dev_server: "npm run dev:admin"
      url: "http://localhost:4000"

See config-reference.md for the full behavior.

ticket_prefix: TICK
tickets_dir: .lanegate/tickets
worktrees_dir: .lanegate/worktrees
executor: claude

core_files:
  - src/core/auth.py
core_patterns:
  - "src/db/**"

verification:
  groups: []   # see above — populate for projects with a UI to visually verify

lock_statuses: [in_progress, code_complete, in_review]

safeguards:
  pre_complete:
    - pytest
  pre_merge:
    - pytest
  post_merge:
    - pytest

flag_file: ~/.lanegate/feature_flags.json

environments:
  - name: staging
    branch: staging
    from: main
    trigger: auto        # lanegate auto-promotes this env during cmd_merge when from matches merge target
    flag_file: ~/.lanegate/staging.feature_flags.json

  - name: production
    branch: production
    from: main
    trigger: manual      # LaneGate promote production runs the full sequence
    guard_script: scripts/check-deploy-window.sh
    pre_promote:
      - scripts/run-integration-tests.sh
    post_promote:
      - scripts/restart-service.sh
    flag_file: ~/.lanegate/feature_flags.json

---

V1 Coordination Model

Scope: single git checkout, single machine.

The touches list in each ticket is a pessimistic file-level lock. LaneGate holds the lock from in_progress through in_review, releasing only at merged. This reduces edit collisions when multiple agents work in parallel.

Agent A: TICK-003  touches: [src/auth.py]    status: in_progress  → LOCKED
Agent B: TICK-007  touches: [src/billing.py] status: in_progress  → LOCKED
Agent C: TICK-009  touches: [src/auth.py]    status: open         → BLOCKED (overlaps A)

What this does NOT prevent: semantic conflicts. If one ticket changes an exported API and another ticket changes a caller in a different file, both tickets may be touch-disjoint and still incompatible. LaneGate relies on safeguards, static checks, and review to catch integration problems before they land.

It also does not prevent two separate git clones on two machines from claiming the same ticket. check_local_not_behind_remote runs on every start to reduce this window, but it does not close it entirely. This is a V1 limitation covered in Known Limitations.

Five concurrency bugs fixed versus the original orchestrator:

Bug	Fix
Lock released at code_complete	Lock held until merged
TOCTOU on start	Re-read ticket from disk immediately before write
Merge worktree leaked	Capture worktree path before nulling the field
Case mismatch on macOS	Worktree dirs always lowercased
Substring dedup in gh-sync	Exact [TICK-N] prefix match

---

Testing

python3 -m pytest tests/ -q

Most tests are fast unit-style checks and mock git-facing subprocess calls where needed to isolate edge cases. tests/test_e2e_lifecycle.py is the real lifecycle integration suite: it creates a temporary git repository with git init, makes real commits, creates a real linked worktree, runs the ticket from draft through done, and verifies actual branch, merge, and worktree state. That suite still runs under the default pytest invocation. Only the model response is stubbed, to avoid nondeterminism and external cost.

The release-artifact gate builds a wheel/sdist in a temporary copy, uses a clean non-editable installation, and exercises the CLI in throwaway repositories:

python -m pip install build
python ci/smoke_release.py

See the release smoke-gate guide for its seven checks and the currently expected first-promotion failure in CI.

---

MCP integration

Add to your MCP client config:

{
  "mcpServers": {
    "lanegate": {
      "command": "lanegate",
      "args": ["mcp"]
    }
  }
}

The server starts on stdio. An AI agent can then call board(), next_ticket(), repo_status(), recent_logs(), orchestrate(dry_run=true), start("TICK-007"), etc. as native tools, with no shell commands required. Run lanegate install-agent-tools to write Claude slash commands and reusable Codex or generic MCP config snippets into the repo. See docs/agent-tools.md for the supported agent surfaces and continuation guarantees.

---

Roadmap

Planned work:

• Dependency-aware scope hints using AST/import information
• Stronger sandboxing for agent runs
• Executor conformance tests
• Distributed ticket coordination

See ROADMAP.md for details.

---

Docs

• docs/demo-walkthrough.md — end-to-end walkthrough: idea → analyze → parallel worktrees → review → merge, using a toy calculator project; includes a dry-run path for users without an executor configured
• docs/troubleshooting.md — FAQ and debug steps: executor hangs, stuck tickets, worktree cleanup, lock files, MCP setup, and more
• SECURITY.md — security policy, reporting vulnerabilities, and what LaneGate does and does not do
• docs/security-model.md — full threat model, trust boundaries, executor permission matrix, MCP trust model, V1 limitations, and safe usage recommendations
• docs/ARCHITECTURE.md — built architecture, module map, and design invariants
• docs/config-reference.md — supported .lanegate.yml keys and defaults
• docs/executor-capabilities.md — capability matrix comparing Claude, Codex, Aider, and Ollama across headless support, prompt transport, local model support, auto-commit behavior, and sandbox status
• docs/v2-interface-boundaries.md — V1.5 layer boundaries (Python core / local API / UI add-on / optional runner), the lanegate api endpoint contract and what's built vs. still design-only, and the Go TUI spike result
• .lanegate.yml.example — annotated example configuration