github cloudposse/atmos v1.214.0-rc.2
on GitHub

pre-release7 hours ago
feat: MCP server integrations — connect Atmos to external MCP servers @aknysh (#2267) ## what
  • Connect Atmos to external MCP servers (AWS, GCP, Azure, custom) and use their tools in atmos ai ask, atmos ai chat, and atmos ai exec
  • Add CLI MCP management commands: atmos mcp list, tools, test, status, restart, start, export
  • Add smart server routing — automatically selects only the MCP servers relevant to the user's question using a lightweight routing call to the configured AI provider
  • Add --mcp flag on all AI commands for manual server selection (supports --mcp aws-iam,aws-billing and --mcp aws-iam --mcp aws-billing), env var ATMOS_AI_MCP
  • Add atmos mcp export to emit .mcp.json for Claude Code / Cursor / IDE integration
  • Add Atmos Auth integration — identity field on server config for automatic credential injection (references identities from the auth section)
  • Add toolchain integration — resolves uvx/npx from .tool-versions before starting servers
  • Add BridgedTool pattern to wrap external MCP tools as native Atmos tools.Tool interface
  • Add human-readable tool names in output (aws-iam → list_roles instead of aws-iam__list_roles)
  • Add tool execution display to atmos ai ask output via MarkdownFormatter
  • Add configurable ai.max_tool_iterations (default 25, was hardcoded 10) to support complex multi-tool queries
  • Add complete example with 8 pre-configured AWS MCP servers at examples/mcp/
  • Add comprehensive documentation: MCP Configuration, MCP Commands, AI Landing Page

why

  • Leverage the ecosystem — 100+ MCP servers exist for AWS, GCP, Azure, databases, monitoring, CI/CD. Instead of reimplementing cloud provider functionality, Atmos orchestrates existing MCP servers
  • Parity with AI tools — Claude Code, Cursor, Windsurf all manage MCP servers. Atmos should too
  • Speed — Installing an AWS MCP server takes seconds. Building equivalent functionality takes weeks
  • Composability — Users can mix native Atmos tools (describe stacks, validate) with external tools (AWS billing, security, IAM) in the same AI conversation

references

See It in Action

All outputs below are from real AWS accounts. Account IDs, resource identifiers,
and internal names have been redacted. Cost figures represent an example of real-world spending.

List configured servers

$ atmos mcp list
       NAME         STATUS                           DESCRIPTION
─────────────────────────────────────────────────────────────────────────────────────────
 aws-api            stopped  AWS API — direct AWS CLI access with security controls
 aws-billing        stopped  AWS Billing — billing summaries and payment history
 aws-cloudtrail     stopped  AWS CloudTrail — event history and API call auditing
 aws-docs           stopped  AWS Documentation — search and fetch AWS docs
 aws-iam            stopped  AWS IAM — role/policy analysis and access patterns
 aws-knowledge      stopped  AWS Knowledge — managed AWS knowledge base (remote)
 aws-pricing        stopped  AWS Pricing — real-time pricing and cost analysis
 aws-security       stopped  AWS Security — Well-Architected security posture assessment

Explore tools from a server

$ atmos mcp tools aws-security
           TOOL                                                         DESCRIPTION
──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
 CheckSecurityServices     Verify if selected AWS security services are enabled in the specified region and account.
 GetSecurityFindings       Retrieve security findings from AWS security services.
 GetStoredSecurityContext  Retrieve security services data that was stored in context from a previous CheckSecurityServices call.
 CheckStorageEncryption    Check if AWS storage resources have encryption enabled.
 ListServicesInRegion      List all AWS services being used in a specific region.
 CheckNetworkSecurity      Check if AWS network resources are configured for secure data-in-transit.

Test server connectivity

$ atmos mcp test aws-docs
✓ Server started successfully
✓ Initialization handshake complete
✓ 4 tools available
✓ Server responds to ping

Ask AI — documentation search (smart routing selects aws-knowledge)

$ atmos ai ask "How do I configure S3 bucket lifecycle rules?"
ℹ MCP routing selected 1 of 8 servers: aws-knowledge
ℹ MCP server "aws-knowledge" started (6 tools)
ℹ Registered 6 tools from 1 MCP server(s)
ℹ AI tools initialized: 16
👽 Thinking...

   Configuring S3 Bucket Lifecycle Rules

   S3 lifecycle rules automate object management by transitioning objects between
   storage classes, archiving, or expiring them...

  ## Tool Executions (1)
  1. ✅ aws-knowledge → aws.search_documentation (2874ms)

Ask AI — billing analysis (smart routing selects aws-billing)

$ atmos ai ask "Show our billing summary for the past 2 months"
ℹ MCP routing selected 1 of 8 servers: aws-billing
ℹ MCP server "aws-billing" started (25 tools)
ℹ Registered 25 tools from 1 MCP server(s)
ℹ AI tools initialized: 35
👽 Thinking...

  ## 📊 AWS Billing Summary — February & March 2026

   Service                                  │ Feb 2026 │ Mar 2026 │ Change
  ──────────────────────────────────────────┼──────────┼───────────┼──────────
   Amazon Virtual Private Cloud             │ $309.53  │ $261.17   │ ▼ $48.36
   EC2 - Other                              │ $88.12   │ $123.39   │ ▲ $35.27
   ...

  ## Tool Executions (1)
  1. ✅ aws-billing → cost-explorer (381ms)

Ask AI — security posture across all regions (smart routing selects aws-api + aws-security)

$ atmos ai ask "Is GuardDuty enabled in all regions?"
ℹ MCP routing selected 2 of 8 servers: aws-api, aws-security
ℹ MCP server "aws-api" started (2 tools)
ℹ MCP server "aws-security" started (6 tools)
ℹ Registered 8 tools from 2 MCP server(s)
ℹ AI tools initialized: 18
👽 Thinking...

  ## ❌ GuardDuty is NOT Enabled in All Regions

   GuardDuty is only enabled in 1 out of 34 regions checked.

  ### ✅ Enabled (1 region)

   Region    │ Detector ID
  ───────────┼────────────────────────────────────
   us-east-2 │  <detector-id-redacted>

  ### ❌ Not Enabled (33 regions)

   Region         │ Region         │ Region
  ────────────────┼────────────────┼────────────────
   ap-south-1     │ ap-south-2     │ ap-southeast-1
   eu-central-1   │ eu-west-1      │ us-east-1
   us-west-2      │ ...            │

  ### 🔒 Recommendations
    1. Enable GuardDuty in all active regions
    2. Use delegated administrator via AWS Organizations
    3. Prioritize us-east-1, us-west-2, eu-west-1 immediately

  ## Tool Executions (4)
  1. ✅ aws-api → call_aws (400ms)
  2. ✅ aws-api → call_aws (14ms)
  3. ✅ aws-api → call_aws (7ms)
  4. ✅ aws-api → call_aws (9450ms)

Ask AI — IAM audit (smart routing selects aws-iam)

$ atmos ai ask "List all IAM roles with admin access"
ℹ MCP routing selected 1 of 8 servers: aws-iam
ℹ MCP server "aws-iam" started (29 tools)
ℹ Registered 29 tools from 1 MCP server(s)
ℹ AI tools initialized: 39
👽 Thinking...

  ## 🔐 IAM Roles with Admin Access

  ### 1. ✅ Direct AdministratorAccess Policy (4 attachments)

   Role Name                                        │ Description                                    │ Trust Principal
  ──────────────────────────────────────────────────┼────────────────────────────────────────────────┼───────────────────────────
    AWSReservedSSO_AdministratorAccess_...          │ Allow Full Administrator access to the account │ AWS SSO (SAML Federation)
    AWSReservedSSO_RootAccess_...                   │ Centralized root access to member accounts     │ AWS SSO (SAML Federation)
    AWSReservedSSO_TerraformApplyAccess_...         │ Full Terraform state and account access        │ AWS SSO (SAML Federation)
    AWSReservedSSO_TerraformApplyAccess-Core_...    │ Full Terraform access (core backend)           │ AWS SSO (SAML Federation)

  ### 🛡️ Security Recommendations
    1. Review SSO assignments for AdministratorAccess and RootAccess roles.
    2. Audit TerraformApplyAccess roles — ensure MFA/session policies are enforced.
    3. Monitor tfstate roles — cross-account trust across 14 accounts.
    4. Enable CloudTrail for AssumeRole calls on high-privilege roles.

  ## Tool Executions (2)
  1. ✅ aws-iam → list_roles (314ms)
  2. ✅ aws-iam → list_policies (174ms)

Check status of all servers

$ atmos mcp status
      NAME       STATUS   TOOLS                        DESCRIPTION
─────────────────────────────────────────────────────────────────────────────────────────
 aws-api         running  2      AWS API — direct AWS CLI access with security controls
 aws-billing     running  25     AWS Billing — billing summaries and payment history
 aws-cloudtrail  running  5      AWS CloudTrail — event history and API call auditing
 aws-docs        running  4      AWS Documentation — search and fetch AWS docs
 aws-iam         running  29     AWS IAM — role/policy analysis and access patterns
 aws-knowledge   running  6      AWS Knowledge — managed AWS knowledge base (remote)
 aws-pricing     running  9      AWS Pricing — real-time pricing and cost analysis
 aws-security    running  6      AWS Security — Well-Architected security posture assessment

Summary by CodeRabbit

  • New Features

    • External MCP server support with new mcp commands: list, tools, test, status, restart, export
    • New --mcp / ATMOS_AI_MCP flag for ai ask/chat/exec to select servers (overrides routing)
    • Smart MCP routing to choose relevant servers per prompt
    • Human-friendly tool names in AI responses
    • Configurable AI request timeouts and max tool‑iteration limits

  • Documentation

    • Extensive MCP and AI integration docs, examples, and an AWS MCP example

🚀 Enhancements

fix: add process-level credential cache @AleksandrMatveev (#2272) ## what
  • Added a process-level in-memory credential cache (sync.Map) to authenticateChain() that stores successfully authenticated credentials keyed by realm + chain identity
  • When a subsequent authentication request matches the same realm and chain within the same CLI invocation, the cached credentials are returned (after expiration validation) without making additional API calls
  • The previous fix (dbcba35) that skips cached target identity credentials in keyring/file storage remains intact - this cache layer sits above it

why

  • The previous fix correctly prevented stale cached credentials from being returned by always forcing re-authentication of the target identity (e.g., AssumeRole). However, during atmos describe affected, each !terraform.state YAML function resolution creates a new AuthManager via resolveAuthManagerForNestedComponent, and each one triggers a full AssumeRole API call
  • This caused atmos describe affected to degrade from ~2 minutes to ~17 minutes due to N redundant STS AssumeRole calls for N components sharing the same auth chain
  • The in-memory cache is inherently safe: unlike keyring/file caches that persist across processes and may contain stale data from different auth mechanisms (e.g., pod credentials vs. role credentials), process-scoped credentials were authenticated during the current invocation and are guaranteed correct
  • Cached entries are validated against the existing expiration buffer (minCredentialValidityBuffer = 15m) before reuse, and the cache resets naturally when the process exits

references

Summary by CodeRabbit

  • New Features

    • Process-level in-memory credential cache to speed repeated authentications and share valid credentials across instances.
    • Automatic detection and removal of expired/invalid cached credentials with transparent re-authentication.
    • Ability to reset the process credential cache.

  • Tests

    • Tests verifying cache hits, isolation by realm/chain, expiration handling, and cache reset behavior.
fix: use allowlist DTO for instance uploads to prevent sensitive data leakage @milldr (#2269) ## What - Introduce `dtos.UploadInstance` as an allowlist struct with only the fields Atmos Pro needs - Convert `schema.Instance` → `UploadInstance` at the upload boundary in `list_instances.go` - Sanitize `settings.pro` to handle `map[interface{}]interface{}` from YAML for JSON compatibility

Fields included in upload (allowlist)

  • component — instance identification
  • stack — instance identification
  • component_type — terraform or helm
  • settings.pro — drift detection configuration (enabled flag, detect/remediate workflows)

Fields excluded (never serialized)

  • vars — can contain secrets
  • env — can contain secrets and credentials
  • backend — contains role ARNs and bucket names
  • source — not used by Atmos Pro
  • metadata — not used by Atmos Pro
  • All settings keys except pro — not used by Atmos Pro

Why

  • Previously, InstancesUploadRequest used []schema.Instance which included all sections from the stack config
  • env and vars can contain secrets that should never leave the CI environment
  • Nested YAML maps in the excluded fields produce map[interface{}]interface{} types that encoding/json cannot marshal, causing atmos list instances --upload to fail with json: unsupported type: map[interface {}]interface {}
  • Using an allowlist DTO ensures new fields added to schema.Instance are never inadvertently uploaded

Summary by CodeRabbit

Release Notes

  • Refactor
    • Optimized instance upload payloads by streamlining the data structure to include only essential configuration fields (component, stack, component type, and settings).
    • Enhanced upload efficiency by automatically filtering and extracting pro-specific settings before transmission.
refactor(auth): move ECR login from auth to aws namespace (ATMOS-37) @Benbentwo (#2144) ## what
  • Move atmos auth ecr-login command to atmos aws ecr login under the AWS namespace
  • Create new cmd/aws/ecr/ package with parent ECR command and login subcommand
  • Move ECR login tests to new package structure (16 tests, all passing)
  • Relocate documentation from auth/ to aws/ command directory
  • Update all cross-references in tutorials, blog posts, and internal design docs

why

The auth namespace must remain provider-agnostic per CLI design principles. AWS-specific commands like ECR login belong under the atmos aws namespace hierarchy, following the established pattern with atmos aws eks update-kubeconfig. This ensures the auth namespace is not polluted by provider- or service-specific commands and maintains a clean separation between generic auth operations and cloud-specific integrations.

references

Closes #ATMOS-37

Acceptance criteria from ATMOS-37:

  • ✅ No AWS- or ECR-specific commands exist directly under atmos auth
  • ✅ Command structure aligns with interface-based design
  • ✅ CLI help and docs reflect the updated command hierarchy

Summary by CodeRabbit

  • New Features

    • Moved ECR login into the AWS namespace as atmos aws ecr login, added top-level ecr subcommand, introduced a command-local --identity flag, and retained multi-registry --registry support.

  • Tests

    • Removed legacy test suite and added a comprehensive test suite validating the new command wiring, flags, argument flows, and auth-manager behavior.

  • Documentation

    • Updated CLI docs, tutorials, PRD, and blog examples to reference atmos aws ecr login.

  • Chores

    • Added new sentinel errors for clearer ECR login failure and identity-selection reporting.
Check out latest releases or
releases around cloudposse/atmos v1.214.0-rc.2

Don't miss a new atmos release

NewReleases is sending notifications on new releases.

Get notifications