Open source · MIT · Built by Hackit
Slick writes its own skills out of the work it just did, improves them while it uses them, and builds a model of who you are that deepens every session. It is not tied to your laptop — run it on a $5 VPS, a GPU cluster, or serverless infrastructure that costs almost nothing while idle.
curl -fsSL https://slick-agent.hackit.cc/install.sh | bashiex (irm https://slick-agent.hackit.cc/install.ps1)Skills · memory · sessions · subagents — converging
Most agents are a very good one-shot. Every session opens on an empty desk — the procedure you worked out together yesterday is gone, and you type it again. Slick closes that gap on its own, and the loop is the product.
You hand it a task. It runs tools, spawns isolated subagents, and writes Python that calls those tools over RPC — collapsing a multi-step pipeline into one turn.
After anything complex it writes the procedure down as a skill. Nobody asks it to. The next run starts from the method, not from scratch.
Skills are edited while they are used, not between releases. A skill that misfires once comes back corrected.
FTS5 search over every past session, summarized on demand, plus periodic nudges to persist what mattered — and a dialectic model of who you are that deepens each time you talk.
Multiline editing, slash-command autocomplete, history, interrupt-and-redirect mid-run, and tool output that streams as it happens. Then close the laptop — the same conversation is waiting in Telegram.
A full TUI: multiline editing, slash-command autocomplete, conversation history, interrupt-and-redirect, and streaming tool output.
Telegram, Discord, Slack, WhatsApp, Signal and the CLI, all from one gateway process — with voice memo transcription and continuity across every one of them.
Agent-curated memory, autonomous skill creation, skills that self-improve in use, cross-session recall, and Honcho user modeling. Compatible with the agentskills.io standard.
A built-in cron scheduler that delivers to any platform. Daily reports, nightly backups, weekly audits — described in plain language, running unattended.
Spawn isolated subagents for parallel workstreams, each with its own context, reporting back into yours.
Batch trajectory generation and trajectory compression, for training the next generation of tool-calling models.
Daytona and Modal give the agent serverless persistence: the environment hibernates when idle and wakes on demand, costing next to nothing between sessions. Or point it at a $5 VPS and leave it there.
Point Slick at whichever provider you already pay for. Switching is one command — no code changes, no migration, no rewriting your skills.
slick modelInstall in one line
curl -fsSL https://slick-agent.hackit.cc/install.sh | bash