For teams building with AI agents
Decisions that outlive the session.
Reconry gives AI agents the approved decisions, project rules, and constraints they need before they start work. When a plan conflicts with something your team already decided, Reconry catches it and brings the change back to a person.
Replace Supabase Auth with Auth0.
Plan written by Claude Code. No person has reviewed it.
This plan conflicts with an approved decision. The team chose Supabase Auth for this project.
Authentication stays on Supabase Auth. RLS policies are written against auth.uid() throughout.
Replace Supabase Auth with Auth0.
A person must approve this change. If the agent proposes a different direction, the proposal goes to someone with approval rights.
Nothing was changed.
Works anywhere your agents work
Reconry sits underneath the tools you already use.
What Reconry is
Give every agent the decisions that matter.
Project decisions are often scattered across documents, tickets, conversations, and code. Reconry brings those decisions together, shows which ones have been approved, and gives each agent the ones that apply to its work.
This is decision authority: a clear, human-approved record of what governs the work.
- Decision
- Authentication stays on Supabase Auth.
- Project rule
- Every table carries a row-level security policy.
- Constraint
- Session tokens expire after 30 minutes of inactivity.
- Known issue
- Bulk imports over 5,000 rows time out in edge functions.
- Approved exception
- The admin export may skip pagination until 1 Dec 2026.
How it works in practice
Where Reconry helps
Reconry helps agents start with the right information, follow the decisions your team has already made, and bring conflicts back to people before work moves forward.
Give agents the decisions that matter before they begin
Agents often start work with a ticket, prompt, or document but miss important decisions made somewhere else.
What Reconry does
Reconry gives the agent the approved decisions, project rules, and limits that apply to the work.
What your team controls
Your team decides what counts as an approved project decision. A document or message does not become official just because Reconry found it.
Agents propose. People approve. Nothing changes automatically.
From project to agent-ready
See Reconry in action
Start with the project you already have. Reconry helps your team turn scattered information into clear decisions that agents can follow.
Bring in the project you already have
Connect the documents, tickets, plans, and decisions your team already uses. Reconry organizes what it finds into a draft project view.
Stays with your team
Finding something does not make it approved. Your team confirms what has actually been decided.
Found is not the same as approved. Only your team confirms a decision.
Give agents the decisions that matter
Before work moves forward, Reconry gives the agent the approved decisions and project rules that apply to the task.
If the proposed plan goes against something your team already decided, Reconry clearly shows the conflict.
Stays with your team
The agent can explain the conflict and suggest a change, but it cannot approve one.
Review changes and keep the project aligned
Your team reviews conflicts, open questions, and suggested changes in one place.
Approved changes become part of the project record and can guide future agent work.
Stays with your team
Nothing changes automatically. People decide what the project should follow.
Reconry does not edit your documents, tickets, or plans.
Reconry does not do the agent’s work, and it does not replace human review. Agents suggest. People approve. Nothing changes automatically.
Integrations
Works where your agents work.
Connect Reconry to the tools your team already uses. Each agent can ask which decisions apply to its task and check a proposed plan before work begins.
Agents
Connects over MCPThese tools support MCP, the open standard Reconry uses to connect. Confirmed support for each tool is listed on its integration page.
External systems
Planned · Not yet availableSystems such as GitHub, so Reconry can read what is in a repository and compare it with what the team approved. Each project grants access one resource at a time.
Common questions
Reconry helps AI agents follow the decisions your team has already made. It keeps your approved decisions, project rules, and constraints in one place. Before an agent starts work, Reconry gives it the ones that apply to its task. If the agent’s plan conflicts with a decision, Reconry points it out and sends any proposed change to a person. Agents propose. People approve.
Decision authority is the clear, human-approved record of what governs a project: its decisions, rules, constraints, known issues, and approved exceptions. A statement in a document is not an approved decision until a person approves it. Reconry keeps who approved each one, when, and why.
Shared context tells an agent what exists. It does not tell the agent which statement to follow when two sources disagree. A repository, a prompt file, and an old specification can each say something different. Reconry tells the agent what the team actually approved, before work begins.
Memory stores information. A knowledge base organizes it. Neither tells an agent which statement the team approved. Reconry keeps only what a person approved, records who approved it and why, and gives each agent the few decisions that apply to its task instead of every project document.
Reconry checks agent plans against your approved decisions. It points out three kinds of issue. A conflict means two statements or actions cannot both be followed. Drift means the project has moved away from an approved decision. An open question means the work needs a decision nobody has made. Each issue quotes the exact passages involved. Checking finished code directly is a planned capability and not yet available.
No. An agent can read the decisions that apply to its task, propose a change, or ask for an exception. It cannot approve its own change or exception, approve work, or grant access. Every proposed change waits for a person with approval rights. Nothing changes automatically.
No. CLAUDE.md, AGENTS.md, ADRs, and specifications stay where they are. Reconry reads them, finds the decisions and rules inside them, and asks your team which ones agents should follow. Each approved decision keeps a link to the passage it came from.
Reconry connects to agents over MCP, an open standard many AI coding tools support. Tools such as Claude Code, Codex, Cursor, and Replit can connect this way. Connections to external systems such as GitHub are planned and not yet available.
Yes. Bring in the documents behind an existing project. Reconry gathers the decisions it finds, quotes each one with its file and line, shows where sources conflict, and lists questions the documents do not answer. Your team approves, edits, skips, or saves each item for later. Reading connected repositories directly is planned and not yet available.
A Project Blueprint is the review Reconry prepares when you bring in a project. It lists the sources Reconry read, the decisions it found, conflicting statements, and open questions. Nothing in it becomes official until a person approves it. It is not a score, and a skipped item is simply skipped.
Start with one project.
Connect the documents and repositories behind an existing project. Reconry will organize the decisions it finds so your team can review what applies before the next agent starts work.