anecticoConnect
Use the interface that fits a coding session or a terminal: MCP, the CLI or both.
Agents & MCP
Your agent does the work.You open the proof.
Connect Claude Code, Codex, Cursor or Gemini with one command. Your agent searches, measures and sets things up through MCP and the CLI. You keep the keys.
Example: an agent session and its proofMCP and CLI / Proof links / Scoped keys
- Bring your own agent
Anectico is the platform, not the agent. The model and the skills stay yours.
- Give it useful evidence
It searches the connected person graph and follows source references.
- Keep access deliberate
Scope the key, review changes and revoke access when you need to.
01One command
Connect once. Ask from then on.
Sign in with “anectico login”, then run “anectico agent bootstrap --project <project-id> --host claude”. It creates a narrow key for one project, sets up your host and checks the connection. Then ask your agent to list your projects.
- Choose claude, codex, cursor or gemini as the host.
- The key is read-only, belongs to one project and expires.
- The secret goes to your keychain, not your terminal history.
02Control
You hold the keys.
The Console is where you control access. See every agent and key, and revoke either. Stop an agent with a request from one person and an approval from a different one. A change your agent makes previews first and applies after you confirm.
- Agents and keys: list and revoke approvals and keys.
- Agent safety: one person asks, a different person approves.
- Audit log: see who did what.
The workflow, connected
Connect. Ask. Open the proof.
Your agent finds its way through the product without a dashboard to browse.
Connect your agent
Run one command, or point any MCP client that supports Streamable HTTP and a fixed header at the URL with a key.
Ask your question
Your agent discovers its tools, finds the evidence and fetches the references it follows.
Open the proof
Check the link your agent gives you. Review a preview before any change, and revoke access when the work ends.
The capabilities behind it
Tools, evidence and control.
Your agent gets discoverable tools and connected evidence. You get scoped access, confirmation and an audit trail.
Work with the evidence
Your agent finds the relevant sources, follows the evidence and ends with a link you can open.
Govern the work
Bound each key. Keep confirmation, audit, stop and revocation in the workflow.
Put it to work
Bring the production question into your coding session.
Why did this customer’s checkout fail?
Ask your agent to find the customer, read the related request and give you the link.
Explore this workflowWhat changed around this error?
It reads the Issue, the telemetry and the release context that its key allows.
Explore this workflowCan the agent help with incident work?
Give it the response scopes for the job. Changes preview first.
Explore this workflowYour first useful workflow
Start with the client you already use.
The connection guide covers each host. Begin with read access, then add scopes for the job.
Choose the connection
Run the bootstrap command for your host, or follow the MCP guide.
Read the guide
A few useful answers
A few useful details.
Which agents can I use?
One command sets up Claude Code, Codex, Cursor and Gemini. Any MCP client that supports Streamable HTTP and a fixed header can connect with a URL and a key. The compatibility page lists, host by host, what has been run and what has not.
Does Anectico have an assistant or a copilot?
No. Anectico is the platform, not the agent. There is no chat, no copilot and no automatic fixing. You bring your own agent, model and skills.
How are changes kept under control?
A key has one project and a list of scopes. A change previews first and applies only after a second call with a confirmation token. You can revoke any agent or key in the Console.
Is this the same as monitoring my own AI agent?
No. This page is about using an agent to work with Anectico. AI & agent observability covers the calls, runs, tools and quality evidence produced by your own application.
Does action confirmation mean a person approved?
No. The server’s preview-and-confirm step and your MCP host’s own approval settings are separate. Configure and verify the host review behavior you want.
Is the Console required for every task?
No. Your agent works through MCP and the CLI. You use the Console to control access and to open proof pages. Sign-in and workspace membership must still be set up.
Anectico / Private preview
Bring your agent.Give it connected context.
Connect your agent and ask your first question in Anectico private preview.