Ferryman automates research, content, and operations on your local machine. Your data never leaves. Your API keys, your models, your rules.
These are real workflows running in production today. Each one is powered by a Skill — a reusable playbook that Ferryman executes end-to-end.
Automatically research keywords, analyze competitor rankings, scrape SERP intent, and generate a structured content calendar with topic clusters — all saved as local Markdown files.
Every morning at 8:00, Ferryman scans 20+ AI sources, filters noise, extracts key signals, and generates a structured intelligence report on your desktop.
Discover free, high-authority backlink opportunities. Ferryman browses directories, forums, and resource pages, filters by domain authority, and compiles a submission-ready report.
Monitor market signals or brand mentions, distill noise into risk scores, and generate automated alerts.
Autonomously repurpose source content for LinkedIn, Twitter, and Newsletters while maintaining brand voice.
A self-directed explorer for emerging niches, competitor pricing models, or technology shifts.
Ships with 6 built-in Skills. While Skills provide optimized playbooks, Ferryman is a general-purpose agent capable of autonomous reasoning for any open-ended task.
Ferryman does not stop at a chat answer. It writes a clean local report with the summary, signals, and next actions already organized.
GeoSolver already has traction on alternative-intent pages. The clearest upside is educational content, GeoGuessr strategy guides, and richer structured SEO assets that connect top-of-funnel traffic to tool pages.
Three commitments that shape every decision we make.
You shouldn't need to write code to automate your work. Ferryman wraps complex toolchains — browser control, file I/O, LLM orchestration — behind a conversational interface with a visual task monitor. If you can describe what you want, you can run it.
Every tool call is logged. Every task has a visible state. Every output is saved to disk. When automation fails — and it will — you can diagnose from structured logs instead of guessing. No black boxes, no invisible magic.
Your data stays on your machine. Your API keys connect directly to your chosen model provider — we never proxy, log, or store them. All artifacts, logs, and configuration live under ~/.ferryman/. BYOK is not optional, it's the architecture.
Full transparency on the systems that power Ferryman. No hand-waving.
Each Session owns an isolated workspace directory, conversation history, and a compressed memory summary. Context never bleeds across projects. Long conversations stay coherent through hybrid memory — recent messages in a sliding buffer, older context in a rolling summary.
Every tool call emits a structured event with start/complete/error phases and millisecond-level duration. Tasks follow a state machine (pending → running → success/failed). Token usage is tracked per session.
Five built-in toolkits: a stealth browser with page distillation, a sandboxed file system, a skill-scoped script runner, a task tracker, and a skill delegation engine. Each runs inside the session workspace with strict path isolation.
Schedules persist in SQLite with cron expressions and track last/next run times. A daily AI hotspot scan or weekly SEO audit keeps producing deliverables without you watching.
A Skill is a local directory: SKILL.md (SOP + YAML metadata), optional scripts, reference materials, and output templates. It can be zipped, shared across teams, or published with a single tool call. Automation experience becomes a portable, versionable asset — not a fragile prompt.
Different work deserves different engines. Ferryman routes planning, execution, and formatting to the model path that fits the job. You choose the providers, configure the keys, and control the routing.
Ferryman is open source on GitHub and runs on macOS. Download the latest release and bring your own API key to get started.
Download for macOS