Longshot — Structured AI-driven development

A structured workflow for AI-driven development

Specs drive the work. Diffs prove the result. Every change is verified and auditable.

Diff review and approval

See exactly what the agent changed — full diffs with file list, change summary, and one-tap approve or reject. Every approved change is a clean, scoped commit.

Verification results with pass and fail (desktop)

Verify before you ship

Run tests, capture screenshots, or write custom checks — all before approving. See pass/fail results with full artifacts. Evidence that the work meets your criteria before it ships.

Full agent transparency

See every file read, code edit, shell command, and test run as it happens. Full visibility into what the agent is doing and why — no black box.

Spec-driven task queue

Queue up tasks with structured specs. Agents process them in isolated worktrees — parallel work without conflicts. See status at a glance: drafting, in progress, verifying, complete.

Task detail with agent-drafted spec (desktop)

Agent-researched specs

Describe a task in a sentence. The agent researches your codebase — reads files, understands architecture, finds relevant patterns — then writes a structured spec with scope, approach, and acceptance criteria. You review and refine. The agent does the research; you steer.

Chat interface with activity feed (desktop)

Activity feed & services

Live feed of everything happening in your project. Manage dev servers and background processes. Start new tasks from conversation.

See how it works →

Different agents for different jobs

Configure which AI backend handles each stage of the task lifecycle. Use a fast model for spec drafting, your best model for implementation.

Per-stage backend configuration Assign different agents to drafting, refining, implementing, fixing, and verifying. Optimise for speed where it matters, capability where it counts.

Claude Code, Codex, Gemini Supports multiple agent backends as interchangeable providers. Switch between them per-task or per-stage without changing your workflow.

Parallel execution Run multiple agents simultaneously, each in its own isolated worktree. Queue tasks and agents process them in parallel — no conflicts, no waiting.

Cost-efficient by default Route cheap work to cheaper models. A spec draft doesn’t need your most expensive model — save that for the implementation that matters.

Full audit trail for every task

Know exactly what was done, why, and when — for every change in your codebase.

Task specs committed to git Every task description, acceptance criteria, and refinement conversation is stored in .longshot/ and committed alongside your code.

Full agent run history Every file read, code edit, shell command, and test run is logged. Replay exactly what the agent did and why.

Verification artifacts Screenshots, test results, and custom check output are saved per-task. Evidence that the work was verified before approval.

Generate change notes Each approved task is a clean git commit with clear scope. Task specs double as change documentation — perfect for release notes, handoffs, and code reviews.

Get started in seconds

One command to install. Works with any codebase.

curl -fsSL https://longshothq.com/install.sh | sh

Then run longshot in your project directory and open the URL on your phone.

Simple pricing

Start free. Upgrade when you need remote access and multiple agents.

Free

$0

Self-hosted, full power

1 agent connection
Unlimited tasks
Diff review & approval
Verification agents
Git sync & branches
All core features

Install Now

Pro

$5/mo

Mobile access from anywhere

3 agent connections
Managed cloud proxy
Push notifications
Access from any network
All Free features
Priority support

Upgrade to Pro

Clean commits, full audit trail

Free to use. No credit card required.