About AWS IoT Core
AWS IoT Core MCP Server exposes device registry, shadows, jobs, rules, and messaging as tools for teams managing AWS IoT Core devices and fleets.
Key features
- 01Manage devices and groups
- 02Read and update device shadows
- 03Create and track fleet jobs
- 04Publish MQTT messages
- 05Manage certificates, policies, and rules
Use cases
- 01Look up or maintain device records
- 02Update desired or reported device state
- 03Message a fleet or schedule fleet jobs
MCP support
It provides MCP clients with tools for AWS IoT Core device registry, shadows, jobs, rules, messaging, certificates, and policies.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
uvx aws-iot-core-mcp-serverAWS_ACCESS_KEY_IDSecretAWS access key IDAWS_SECRET_ACCESS_KEYSecretAWS secret access keyAWS_SESSION_TOKENSecretAWS session token, for temporary credentialsAWS_PROFILENamed profile from ~/.aws/credentials, as an alternative to explicit keysAWS_REGIONAWS region -- required, AWS IoT Core is regionalLocal servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEAWS IoT Core MCP Server
A Model Context Protocol (MCP) server for AWS IoT Core — device registry, device shadows, jobs (fleet management), the rules engine, and messaging.
Scoped deliberately to IoT Core itself: no Greengrass (separate edge-deployment API), no IoT SiteWise (already has an official awslabs MCP server), no Device Defender/Events/TwinMaker — each of those is its own AWS service with its own API and would be its own connector.
Tools
Things (device registry)
| Tool | Description |
|---|---|
list_things(thing_type_name, attribute_name, attribute_value, thing_group_name, max_results) | List things, optionally filtered. |
search_things(query_string, index_name, max_results) | Fleet-indexing search, e.g. attributes.location:"lab-1". Requires indexing to be enabled. |
get_thing(thing_name) | Fetch a thing's attributes and type. |
create_thing(thing_name, thing_type_name, attributes, billing_group_name, overwrite=False) | Register a new thing. |
update_thing(thing_name, attributes, thing_type_name, remove_thing_type, attributes_merge=True) | Update a thing's attributes/type. |
delete_thing(thing_name, confirm=False) | Delete a thing. Irreversible. |
Thing Types & Thing Groups
| Tool | Description |
|---|---|
list_thing_types(thing_type_name) / get_thing_type(thing_type_name) | List/read thing types. |
create_thing_type(thing_type_name, description, searchable_attributes) | Create a thing type. Immutable once created (AWS limitation) — only deprecation is supported after. |
deprecate_thing_type(thing_type_name, undo=False) | Deprecate (or un-deprecate) a thing type. |
list_thing_groups(parent_group, name_prefix_filter) / get_thing_group(thing_group_name) | List/read thing groups. |
create_thing_group(thing_group_name, parent_group_name, attributes, overwrite=False) | Create a thing group. |
delete_thing_group(thing_group_name, confirm=False) | Delete a thing group. Irreversible. |
modify_thing_group_membership(thing_name, thing_group_name, action) | action="add" or "remove" a thing from a group. |
list_things_in_thing_group(thing_group_name) | List members of a group. |
Device Shadows (data plane)
| Tool | Description |
|---|---|
get_thing_shadow(thing_name, shadow_name) | Read a thing's shadow document (desired/reported/delta). Omit shadow_name for the classic (unnamed) shadow. |
update_thing_shadow(thing_name, desired, reported, shadow_name) | Merge-update desired and/or reported state. |
delete_thing_shadow(thing_name, shadow_name, confirm=False) | Delete a shadow document. |
list_named_shadows_for_thing(thing_name) | List named (non-classic) shadows for a thing. |
Messaging
| Tool | Description |
|---|---|
publish_message(topic, payload, qos=0) | Publish an MQTT message to the device fleet via the IoT Data Plane. |
Jobs (fleet management)
| Tool | Description |
|---|---|
list_jobs(status, thing_group_name, max_results) / get_job(job_id) | List/read jobs. |
create_job(job_id, thing_arns, document, description, target_selection="SNAPSHOT") | Create a fleet job (e.g. an OTA update or remote command). |
cancel_job(job_id, confirm=False, comment, force=False) | Cancel a job. |
delete_job(job_id, confirm=False, force=False) | Delete a job. Irreversible. |
list_job_executions_for_job(job_id, status) / list_job_executions_for_thing(thing_name, status) | Track per-device execution status. |
Certificates & Policies (device auth)
| Tool | Description |
|---|---|
create_keys_and_certificate(set_as_active=True) | Generate a new key pair + certificate. The private key is returned only once — capture it immediately. |
get_certificate(certificate_id) / list_certificates(page_size) | Read certificate metadata. |
set_certificate_status(certificate_id, status) | ACTIVE, INACTIVE, or REVOKED. |
delete_certificate(certificate_id, confirm=False, force_delete=False) | Delete a certificate. Irreversible. |
create_policy(policy_name, policy_document, overwrite=False) / get_policy / list_policies | Manage IAM-style IoT policies (JSON documents). |
delete_policy(policy_name, confirm=False) | Delete a policy. Irreversible. |
modify_policy_attachment(policy_name, target, action) | Attach/detach a policy to/from a certificate ARN. |
modify_thing_principal(thing_name, principal, action) | Attach/detach a certificate (principal) to/from a thing. |
list_thing_principals(thing_name) / list_principal_things(principal) | Look up thing↔certificate associations either direction. |
Rules Engine
| Tool | Description |
|---|---|
list_topic_rules(topic) / get_topic_rule(rule_name) | List/read SQL-based message routing rules. |
put_topic_rule(rule_name, sql, actions, description, sql_version, overwrite=False) | Create or replace a rule. actions is a list of raw AWS action objects (e.g. {"lambda": {"functionArn": "..."}}). |
set_topic_rule_enabled(rule_name, enabled) | Enable/disable a rule without deleting it. |
delete_topic_rule(rule_name, confirm=False) | Delete a rule. Irreversible. |
Misc
| Tool | Description |
|---|---|
get_endpoint(endpoint_type) | Look up this account's IoT Core data endpoint (default iot:Data-ATS). |
Safety conventions
Same as the rest of this repo: create_*/put_* default to overwrite=False (best-effort — the AWS IoT API itself has no conditional-write header for most of these operations, so this is a describe-then-create check, not fully race-free like Ditto's If-None-Match). Every delete_*/cancel_* requires confirm=True.
Configuration
Standard AWS SDK credential resolution — no custom env vars for auth. Set whichever of these your setup uses:
| Env var | Description |
|---|---|
AWS_ACCESS_KEY_ID / AWS_SECRET_ACCESS_KEY / AWS_SESSION_TOKEN | Explicit credentials. |
AWS_PROFILE | Named profile from ~/.aws/credentials. |
AWS_REGION / AWS_DEFAULT_REGION | Required — AWS IoT Core is regional. |
AWS_ENDPOINT_URL | Override the API endpoint (e.g. for LocalStack). boto3 reads this natively. |
Or an IAM role, if running somewhere with one attached (EC2/ECS/Lambda). See boto3's credential docs.
Running locally
uv sync
export AWS_REGION=us-east-1
# plus whichever credential env vars apply to you
uv run awsiotcoremcpserver.py
See mcp.json for ready-to-use client configs (stdio and Docker).
Testing
This repo's design principles call for testing against a real instance rather than mocks. AWS IoT Core has no official local emulator, and LocalStack's IoT support requires a paid plan (no free tier) — so this server is the one exception in this repo: tests run against moto, a mature, free, open-source AWS API emulator (not hand-rolled mocks — it implements real request/response behavior for the actual boto3 calls this server makes).
uv run pytest -v
License notes
No bundled deployment artifacts here (unlike the other servers) since there's no local broker/service to run — AWS IoT Core only exists as an AWS-hosted service.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.