Fully Agentic Developer Platform

Six agents. One platform.
Zero handoffs.

Six purpose-built AI agents author your platform, validate it in your real cloud against your real policies, and reach your developers through chat, the wf CLI or their IDE. Across AWS, Azure and GCP.

Start free →

Free forever · No commitment · Upgrade anytime

Wayfinder's built-in agents

Secure, hardened, validated agents you can put to work today.

Six agents that author plans, scaffold stacks, validate against your real policies and answer questions. Each runs as the calling user under scoped MCP tools, every action audited.

Conversational

Wayfinder Assistant

Chat to your platform from the UI, the wf CLI or your IDE. Ask anything from "list my workspaces" to "validate this Wayfinder.yaml".

wayfinder-assistant
Developer-side

Stack Definition Creator

Reads your repo and generates a Wayfinder.yaml that ships. Turns code into a deployable stack against the catalogue your platform team curates.

stackdefinition-creator
Platform-side

Cloud Resource Plan Editor

Authors infrastructure plans with you. Cloud-aware - knows the difference between an AWS plan, an Azure plan and a GCP plan.

cloudresourceplan-editor
Platform-side

Cloud Resource Plan Tester

Runs the plan in your real cloud against your real policies before it publishes. Catches what review can't.

cloudresourceplan-tester
Platform-side

Kube Resource Plan Editor

Authors the Helm-backed delivery plans that take a developer's app from definition to running workload.

kuberesourceplan-editor
Research

Terraform Module Search

Finds the right Terraform or OpenTofu module across the registry so plans never start from a blank page.

terraform-module-search
Next release · Bring your own agent

Plug your own AI into the platform.

Bring your own system prompt, LLM provider and MCP servers. Custom agents inherit the same identity, tool scoping and audit trail as the built-in six.

  • Your prompt, your model A custom systemPrompt and userPromptTemplate, with provider choice between Anthropic Claude and Google Gemini.
  • Bind any MCP server Reach your own tenant's MCPs by name, or platform-shipped MCPs via wayfinder/NAME - one or many per agent.
  • Per-MCP tool scope all exposes every tool the MCP serves; selected restricts to a named list - same mechanic the built-in agents use.
  • Optional structured output A resultSchema validates and packages results without writing a parser.
  • Same governance Runs as the calling user, audited end-to-end, AI credits metered per turn.
Get notified when this ships →
Next release · Workflows

Event-driven automation across your platform.

YAML in, automation out. Triggers fire on Wayfinder events, schedules, manual runs or webhooks. Tasks call MCPs, HTTP integrations, plans, AI agents or sub-workflows.

Triggers

  • Resource events (e.g. StackInstance created)
  • Schedules (cron)
  • Manual (wf workflows run)
  • Custom events emitted by other workflows
  • Inbound webhooks (e.g. PagerDuty P1)

Actions

  • mcp/<integration>/<action> - Slack, ServiceNow, anything MCP-shaped
  • http/<integration>/<action> - reusable HTTP actions
  • ai-agent/<name> - invoke a built-in or BYO agent
  • plan/Apply, plan/Destroy, plan/PlanAndValidate
  • action/<name> - saved reusable recipes
  • Built-ins: wait, foreach, parallel, branch, store-data
Example On deploy → notify Slack 
apiVersion: workflows/v3
kind: Workflow
metadata:
  name: deploy-notify
spec:
  on:
    - name: new-instance
      resource: StackInstance
      event: Created
  tasks:
    - name: notify
      action: mcp/slack/post-message
      inputs:
        channel: '#deploys'
        message: 'Deployed {{ .resource.name }}'
Example On PagerDuty webhook → agent investigates 
apiVersion: workflows/v3
kind: Workflow
metadata:
  name: p1-investigate
spec:
  on:
    - name: p1-alert
      action: webhook/my-webhooks/pager-duty
      when: '{{ eq .severity "P1" }}'
  tasks:
    - name: investigate
      on: [p1-alert]
      action: ai-agent/p1-investigator
      inputs:
        service: '{{ .on.p1-alert.service }}'

Workflows engine ships in the next release; webhook triggers, the catalogue UI and the event broker follow in the two releases after that. See the action surface on Integrations.

Get notified when this ships →
In your IDE

Your coding agent now drives your IDP.

Wayfinder ships its own MCP server. One command and your coding agent lists resources, validates manifests and runs wf commands under your RBAC. No extra token to hand out. 

$ wf setup mcp
# detects Claude Code, Cursor, VS Code,
# Windsurf or Antigravity and wires it up
  • Claude Code
  • Cursor
  • VS Code
  • Windsurf
  • Antigravity

Every call carrying the calling user's identity, with the same RBAC as wf.

Next release · External MCP fabric

Plug other MCP servers straight into your platform.

Register external MCP servers - AWS Knowledge, GitHub, your own internal services - and Wayfinder routes them through the agent executor with auth handled. See the full coming-soon list on the Integrations page.

Secure Agents

An agent is not a backdoor.

Every agent runs as the calling user via wf login. Each is bound to a named MCP tool set, never "every tool the server exposes". No shadow service account. Every action audited.

  • Runs as the user Same identity, same scope, same RBAC as wf.
  • MCP tool scoping Each agent binds to its MCP servers with a named tool set (all or selected) - never "every tool the MCP exposes".
  • Per-agent identity Agents are scoped at the tenant or workspace level; one agent can't impersonate another.
  • Audited end-to-end Every tool call by every agent run is captured.
  • AI credits, metered per turn Predictable per-tenant cost; you see what the agents use.
  • Anthropic or Gemini Choose your provider per tenant; bring your own key.

Try the agents on your stack.

Free tier. Full platform. Six agents, ready to work.

Start free →

Free forever · No commitment · Upgrade anytime