For AI Agents

Give your AI agents a live Work Graph to act on.

Connect Claude, ChatGPT, or Claude Code to Vindaris over MCP. Your agents read where execution risk is building, draft the check-in or summary, and open the goal or task that answers it, all inside the permissions you set.

Claude, ChatGPT, Claude Code Permissions you control A person approves the plan
Connect an agent to Vindaris
ClaudeOver MCP
Read the Work Graph: Net revenue retention is drifting. Drafted a check-in and opened a follow-up task.
YOU YouApproved
Reviewed the summary and the task. Approved and assigned to the retention owner.
How it works

What an AI agent actually does on Vindaris.

An agent does not open a dashboard. It connects once, then reads, drafts, and acts over MCP whenever you ask. Here is each moment of that loop.

An AI agent reading Work Graph alerts over MCP
Work GraphOnce connected, the agent queries the Work Graph and acts on goals and tasks over MCP, within the permissions you set.Connect over MCP →
Work Graph alert showing where execution risk is building
Work GraphThe agent reads the same alerts your team sees, each one carrying the reason the work has stalled or drifted.Catch slipping work →
An AI summary generated from work activity
Data StudioAsk for a read and the agent generates the AI summary from work activity, with the risks already surfaced.Automate reporting →
A new goal and task aligned to the objective it serves
Alignment CanvasThe agent opens the goal, KPI, or task and aligns it to the objective it serves, then hands it back for sign-off.Set the guardrails →
What AI agents run on

A live Work Graph your agents can read and act on.

Most AI agents are blind to how execution is actually going, so they produce confident output that changes nothing. Vindaris gives an agent a connected read of the work and the permission to act on it, so it reasons from current reality instead of a snapshot pasted into the prompt.

Connect over MCP

Turn on AI Tools and connect Claude, ChatGPT, or Claude Code. The agent queries and acts in Vindaris over MCP without leaving the assistant you already use.

Read execution risk

The agent pulls Work Graph alerts and sees where effort has stalled or come loose from a goal, so it reasons about current execution risk, not a status report typed last week.

Generate check-ins and summaries

Ask the agent where an initiative stands and it writes the AI summary from work activity, with the risks already surfaced, ready for you to review instead of draft.

Create goals, KPIs, and tasks

The agent opens a new goal, defines a KPI, spins up a project, or files the follow-up task straight in the workspace, so a decision becomes action in one step.

Map work to the objective

The agent sets the alignment between an item and the objective it belongs under and leaves a comment with its reasoning, keeping the map current as the plan moves.

Guardrails you set

Every agent acts inside fine-grained permissions. You decide what it can read, what it can change, and what still waits on a person to approve.

One cycle with Vindaris

Connect, read, draft, act, approve.

An agent does not keep office hours. This is the loop it runs whenever you hand it a question about the strategy, from the first connection to the action a human signs off.

Connect

Turn on AI Tools and link the agent

An admin connects Claude, ChatGPT, or Claude Code over MCP. The agent now has a scoped link into the live workspace, with its permissions set before it touches anything.

An AI agent connected to Vindaris over MCP
Read

Pull the current execution risk

The agent reads the Work Graph alerts and the work activity behind the goals in question, so its answer starts from where effort has actually stalled or drifted.

Work Graph alerts the agent reads before acting
Draft

Generate the check-in or summary

Instead of you writing the update, the agent produces the AI summary from that signal, with the risks and what moved already called out.

An AI summary drafted from work activity
Act

Open the goal, KPI, or task

When the read calls for action, the agent creates the work directly in Vindaris and aligns it to the objective it serves, so a decision becomes action in one step.

A new goal and task aligned to the objective it serves
Approve

You sign off what enters the plan

The change surfaces for a person to approve. The agent did the gathering and the drafting. You keep the decision on what changes the plan.

A person approving the change an agent proposed
Why teams run their agents on Vindaris

An agent is only as good as what it can see and change.

An AI agent with no connection to the work is guessing. It can write a tidy OKR or a confident summary, but it has no read on whether the work behind a goal has moved, and no way to act when it has not. The output sounds right and changes nothing, so a person still goes and gathers the real state by hand before any of it is useful.

Vindaris gives the agent both halves. Over MCP it reads the Work Graph, where execution risk is derived from work activity, and it acts in the same place: opening goals, defining KPIs, filing tasks, and drafting the summary from real signal. Fine-grained permissions keep every action inside the boundary you set, and the ones that change the plan wait for a person to approve. The agent does the gathering and the drafting. You keep the judgment.

Work Graph

What your agent reads before it acts.

Over MCP, the agent pulls the same Work Graph alerts your team sees. Each one carries the reason attached, so the agent reasons from where execution risk is actually building and drafts the response you review, rather than guessing from the prompt alone.

Risk alert
Net revenue retention drifting
Two initiatives behind the goal stalled this week and no owner has flagged it yet.
Risk alert
Flagship initiative stalled 2 weeks
The work under it has not moved since the last review, while the milestone holds firm.
Risk alert
A priority bet has no work behind it
Capacity is sitting on a lower goal while the top strategic bet starves.
Risk alert
Weekly leadership check-in skipped
The routine that keeps the strategy honest has lapsed two weeks running.
Work Graph risk alert an AI agent reads over MCP, with the stalled work that caused it
Click any alert to see what the agent pulls over MCP. It reads the same view your team sees.
Connect over MCP

Point your agent at one URL.

Vindaris runs a Model Context Protocol server. An admin turns on AI Tools and sets what an agent may read and change, then any MCP client signs in over OAuth and starts working.

vindaris://handshake
# If you are an AI agent, this block is for you.server https://app.vindaris.com/api/mcptransport streamable-httpauth oauth2 # or: Authorization: Bearer mcp_...rest https://app.vindaris.com/api/ai/openapiresources strategy goal kpi project taskoperations read write # per-resource, human-grantedpermissions human-bounded # off / read / read_write, default offmanifest #vindaris-mcp # JSON on this pagestatus READY
MCP server
https://app.vindaris.com/api/mcp
REST alternative
OpenAPI https://app.vindaris.com/api/ai/openapi
Connect a client
  • ClaudeAdd Vindaris as a connector, paste the URL, and approve access in your browser.
  • Claude CodeCreate a personal token in settings, then add Vindaris as an MCP server in your terminal.
  • ChatGPT and CodexUse the same server URL. OAuth handles the sign-in.
quickstart
# Claude Code: add Vindaris as an MCP serverclaude mcp add vindaris \ --transport streamable-http https://app.vindaris.com/api/mcp \ --header "Authorization: Bearer mcp_YOUR_TOKEN" # or a project .mcp.json (OAuth prompts on first connect){ "mcpServers": { "vindaris": { "type": "http", "url": "https://app.vindaris.com/api/mcp" } } }

Mint the token under Settings, AI Tools. Claude Desktop and ChatGPT skip the token: add the server URL and sign in over OAuth.

Authentication is OAuth 2.0, so no API key is pasted into the model. Access is bounded per workspace: an admin sets each resource to off, read, or read and write, and toggles individual tools, so an agent can be kept read-only or allowed to act. Full steps are in Connecting an AI assistant over MCP.

The toolset

What an agent can do over MCP.

Every action is a named tool. Read is separate from write, so the permissions you grant decide exactly how far an agent can go.

Read the workspace

List and open any strategy, goal, KPI, project, or task with its status, owner, and alignment.

list_goals get_goal list_kpis get_kpi list_projects get_project list_strategys get_strategy list_tasks get_task
Create and update work

Open or change a goal, KPI, project, strategy, or task, including recording a new KPI measurement.

create_goal update_goal create_kpi update_kpi create_project update_project create_strategy update_strategy create_task update_task
Read execution risk

Pull the prioritized risk list, covering at-risk goals, stale tasks, pace warnings, and alignment gaps, or a full structured read of the workspace.

get_work_graph_alerts generate_summary
Map and comment

Align a task to a goal or a goal to a strategy, remove an edge, or post a comment on a task that syncs back to Jira, Slack, or Asana.

align_entities remove_alignment push_comment
Turn input into tracked work

Convert meeting notes, an email, or any text into tasks, goals, or projects, aligned to the right objective.

ingest_content
Nudge and subscribe

Send an update reminder to the owner of an overdue item, or subscribe a Data Studio, dashboard, or heatmap view to risk alerts.

trigger_reminder toggle_risk_bell
Machine-readable markdown → llms.txt → openapi → tool manifest (json) →
Example session

What it looks like in practice.

An illustrative exchange over MCP. The agent reads the risk, drafts the update, and opens the follow-up, then stops for a person.

session.log
Human

Where is the European expansion strategy slipping, and draft the weekly update.

Agent
→ get_work_graph_alerts() ← "Make onboarding self-serve" at risk, Activation rate at 38% of target, 2 stale tasks → generate_summary({ strategy: "Expand into the European enterprise market" }) ← draft returned → create_task({ title: "Unblock self-serve onboarding", align: "goal" }) ← task created, pending human approval

The expansion is on track except self-serve onboarding, where the Activation rate sits at 38% of target and two tasks have not moved in nine days. I drafted the weekly update and opened a follow-up task to unblock it. Both are waiting for your approval.

Example only. The tool names and the human-approval step are real; the data is illustrative.

Pricing

Free forever, including two integrations

Start free with full goal and strategy management. Add two integrations to sync tasks. Pay only when you add more than 4 people, need more integrations, or more AI summaries.

Free
€0/forever

Best for individual contributors, small teams or startups. No credit card required.

  • All features (incl. Alignment Canvas, My Priorities, Heatmap, Data Studio & Task Board)
  • Set up 2 integrations of your choice
  • Invite up to 4 additional members
  • 5 ad-hoc AI risk summaries per month
Sign Up for Free
Most popular
Connect
€10/user/mo

Best for teams, small agencies, or departmental and team management.

  • All in Free
  • All integrations, unlimited sync
  • Unlimited AI risk summaries & alignment suggestions
  • Up to 10 users within one organization
Start with Connect
Pro
€20/user/mo

Best for multiple teams, exec teams, or full-org deployments.

  • All in Connect
  • Unlimited users within one organization
  • Direct human support email, 24h response time
  • Human-led onboarding support
Start with Pro

See full plan comparison →

Keep going

Set your agents up on Vindaris

The skills that draft the plan, and the two settings that decide how an agent connects and what it may change.

Agent skills
AI agent skills for goals
Turn a goal into OKRs, KPIs, or EOS Rocks inside Claude, Cursor, or ChatGPT.
How-to
Connect an agent over MCP
Turn on AI Tools and connect Claude, ChatGPT, or Claude Code to the workspace.
How-to
Fine-grained permissions
Decide what an agent can read, what it can change, and what needs approval.
Further reading

More on the platform your agents act on.

The product surfaces an agent reads and writes through, explained for a human.

AI OKR software →

AI check-ins and risk alerts for OKRs, with status read from work activity, so a goal that is quietly slipping surfaces before the quarter turns.
Strategy execution software →

Run the whole strategy on a live Work Graph that reads execution risk from work activity and flags a stall weeks before a milestone is missed.
Goal tracking software →

See whether every goal is pacing to target, read from the work activity behind it, so a goal that is green but quietly slipping surfaces before the quarter turns.
Frequently asked questions

Common questions, answered

How do AI agents connect to Vindaris?
An admin turns on AI Tools in Vindaris and connects an assistant such as Claude, ChatGPT, or Claude Code over MCP. The MCP server URL is https://app.vindaris.com/api/mcp and sign-in is OAuth, so no API key is pasted into the model. From then on the agent can query the workspace and act in it: reading Work Graph alerts and work activity, generating AI summaries, and creating or updating goals, KPIs, projects, and tasks, all within the permissions it is granted.
What can an AI agent actually do inside Vindaris?
Over MCP an agent can read the execution risk on the Work Graph, generate an AI summary of where a goal or initiative stands, create goals, KPIs, projects, and tasks, align an item to the objective it serves, and leave comments. It works from current work activity rather than a pasted-in snapshot, and the actions that change the plan can be set to require human approval.
What is the difference between Agent Skills and the MCP connection?
Agent Skills are downloadable Markdown files that help an assistant draft the plan: writing OKRs, defining KPIs, or setting EOS Rocks, and they work in any tool without a signup. The MCP connection is how an agent then acts on a live Vindaris workspace, reading risk, generating summaries, and creating or updating the work. The skills draft the plan and the MCP connection runs it.
Can I control what an AI agent is allowed to change?
Yes. Vindaris uses fine-grained permissions, so you decide what an agent can read, what it can change, and what still needs a person to approve. An agent can be given a narrow, read-only view or the ability to act, and the actions that enter the plan surface for human sign-off.

Give your agents something real to act on.

Connect an assistant and your first tool in minutes. Free forever for goal management.

Connect an agent to Vindaris

No credit card | No migration | No minimum contract

© 2026 Vindaris. All rights reserved. Sitemap