Context that outlasts the session.
Important facts, decisions, and tone live in version-controlled files, not scattered chat histories. Agents never start half-blind.
Docs / System Architecture
The operating system for companies run with AI.
Your company knowledge, policies, and decisions should outlast any individual model or agent. Alacrán keeps your durable core in plain files and puts consequential automation behind human gates.
The operating architecture
Three foundational principles.
Human judgment is a system feature.
Agent jobs land as a change you approve, and every agent is told to stop and ask before money, contracts, publishing, deletion, or credentials.
Run the business and build the product.
Operational records (funnel, finances, content) and developer workflows (agent runs, diffs, reviews) live in one governed cockpit.
The portable core
Four plain folders. Nothing trapped.
Alacrán does not hide your company behind a proprietary SaaS database. Your context travels with you across AI models and tools.
definitions/Foundational company facts.
Durable facts about customer segments, team roles, business ontology, and operating rules.
See company structuredocs/decisions/Never lose the why.
Architectural, product, and business decisions preserved with explicit rationale and rejected alternatives.
How decisions workrecords/Operational department data.
Machine-maintained records for finances, sales leads, content schedules, and audience growth.
Explore recordsLocal-first by design
A folder is the product.
No proprietary cloud database is required to understand your own company.
On your machine
Close Alacrán and the local app stops. The files remain readable and useful.
definitions/
Durable facts about the company, customers, team, and how work flows.
notes/
Working material, session notes, summaries, and information still being shaped.
docs/
Decisions and retrospectives, including the reasoning you will want later.
records/
Operational data for pipeline, finances, content, social, and connected departments.
Questions and Answers
Trust is the product.
How is Alacrán different from an agent harness?
Others run agents in isolated loops. Alacrán provides the company memory, department records, and policy gates that agents work for.
Which AI providers can I use?
Alacrán is model-flexible. Use Claude Code, Kiro, Kimi, Antigravity, or Codex (beta) without rebuilding your company context.
What actions require human approval?
Money, pricing, contracts, commitments, publication, irreversible data deletion, and credential changes are stop-and-ask in every agent's rules.
What happens if I stop using Alacrán?
You keep your folder. All context remains readable Markdown, YAML, and git commits on your computer, never locked in an export.