Release 0.4.1: Sidecar 0.9 discovery and manual AWS deployment

.github/workflows/release-pypi.yml

@@ -154,51 +154,3 @@ jobs:
       - name: Publish to MCP Registry
         run: ./mcp-publisher publish
 
-  deploy:
-    name: Deploy AWS
-    needs: publish-pypi
-    environment: aws
-    runs-on: ubuntu-latest
-    timeout-minutes: 45
-    permissions:
-      contents: read
-    concurrency:
-      group: aws-mcp-production
-      cancel-in-progress: false
-    env:
-      AWS_REGION: us-east-2
-      AWS_DEFAULT_REGION: us-east-2
-      IMAGE_TAG: release-${{ github.ref_name }}-${{ github.run_id }}-${{ github.run_attempt }}
-    steps:
-      - uses: actions/checkout@v7
-      - name: Build AWS image
-        run: python3 scripts/build_container.py --target aws --tag "deep-agentic-core-mcp:$IMAGE_TAG"
-      - name: Configure AWS credentials
-        uses: aws-actions/configure-aws-credentials@e1253824e5c10ff9df46874f81ed3ec929e19cfd # v6
-        with:
-          aws-access-key-id: ${{ secrets.AWS_ACCESS_KEY_ID }}
-          aws-secret-access-key: ${{ secrets.AWS_SECRET_ACCESS_KEY }}
-          aws-region: ${{ env.AWS_REGION }}
-      - name: Publish immutable image
-        id: image
-        shell: bash
-        run: |
-          repository_uri=$(aws cloudformation describe-stacks --stack-name deep-agentic-core-mcp-assets --query 'Stacks[0].Outputs[?OutputKey==`RepositoryUri`].OutputValue | [0]' --output text)
-          registry="${repository_uri%%/*}"
-          repository="${repository_uri#*/}"
-          export DOCKER_CONFIG=$(mktemp -d)
-          trap 'docker logout "$registry" >/dev/null 2>&1; rm -rf "$DOCKER_CONFIG"' EXIT
-          aws ecr get-login-password | docker login --username AWS --password-stdin "$registry"
-          docker tag "deep-agentic-core-mcp:$IMAGE_TAG" "$repository_uri:$IMAGE_TAG"
-          docker push "$repository_uri:$IMAGE_TAG"
-          digest=$(aws ecr describe-images --repository-name "$repository" --image-ids "imageTag=$IMAGE_TAG" --query 'imageDetails[0].imageDigest' --output text)
-          echo "uri=$repository_uri@$digest" >> "$GITHUB_OUTPUT"
-      - name: Deploy image through CloudFormation
-        env:
-          IMAGE_URI: ${{ steps.image.outputs.uri }}
-        run: python3 scripts/aws_ci_deploy.py --image "$IMAGE_URI" --change-set "ci-$GITHUB_RUN_ID-$GITHUB_RUN_ATTEMPT"
-      - name: Verify hosted MCP
-        env:
-          MCP_API_KEY: ${{ secrets.MCP_SMOKE_KEY }}
-          MCP_ORIGIN: https://mcp.deepagentlabs.io
-        run: python3 scripts/ci_smoke.py

AGENTS.md

@@ -130,9 +130,7 @@ Test transport-specific authorization and state isolation when changing dispatch
 - Pull requests, pushes to `main`, and manual CI runs perform checks and package
   builds without deploying AWS or publishing to PyPI.
 - Pushing a `v*` tag runs `release-pypi.yml`: the full test matrix and packaging
-  checks, PyPI publication, GitHub Release, MCP Registry publication, and AWS
-  deployment. AWS depends on successful PyPI publication and deploys the tagged
-  source using an immutable image. Infrastructure changes use the operator flow.
+  checks, PyPI publication, GitHub Release, MCP Registry publication. AWS deployment is manual from 0.4.1. Infrastructure changes use the operator flow.
 - Prepare matching versions in `pyproject.toml`, `__init__.py`, `server.json`,
   and `uv.lock`, plus dated changelog notes, before pushing the release tag.
 - Keep Docker files under `deploy/docker/` and infrastructure under `deploy/aws/`.
@@ -206,8 +204,17 @@ from the same tag and does all of the following:
 - publishes the package to PyPI via Trusted Publishing (OIDC)
 - creates the GitHub Release object for `vX.Y.Z`
 - publishes the server to the MCP Registry
-- deploys the tagged source to AWS after PyPI publication succeeds
+- leaves AWS deployment to an explicit manual operator action
 
 The GitHub Release title is the tag name, and its body is copied from the
 matching `CHANGELOG.md` section so the changelog, tag, PyPI release,
 GitHub Releases page, and MCP Registry publish all stay aligned.
+
+
+## Release pipeline from 0.4.1
+
+Tag releases validate, publish PyPI/GitHub Release and publish the MCP Registry.
+Automatic AWS deployment is removed from GitHub Actions. Existing AWS build and
+operator deployment scripts remain available for an explicit manual deployment;
+publishing a package does not update the running AWS service. Standalone image
+recipes include Sidecar 0.9.0 for package discovery, with no live runtime bridge.

CHANGELOG.md

@@ -14,6 +14,16 @@ This project follows [Semantic Versioning](https://semver.org/).
   dispatch, stop/revoke, denial evidence, hosted Tower credentials) move to
   Phase 5.
 
+## 0.4.1 - 2026-10-19
+
+- Discover installed Sidecar 0.9 task-authority, revocation and redacted-evidence
+  APIs without constructing a task or exposing runtime controls.
+- Pin Sidecar 0.9.0 in standalone hosted image recipes; native evidence remains
+  compatible and no new remote authority API is claimed.
+- Remove automatic AWS deployment from the release pipeline. Package/GitHub/MCP
+  Registry publication remains; existing deployment scripts are manual operator tools.
+- Document Sidecar authority, privacy, evidence and ownership boundaries.
+
 ## 0.4.0 - 2026-10-19
 
 First Phase 4 (Unified Workflows) slice. No sibling release, dependency range

CONTRIBUTING.md

@@ -69,24 +69,17 @@ Never commit credentials, generated API keys, or local deployment state.
 Pull requests, pushes to `main`, and manual CI runs perform lint, format, typing,
 tests across Python 3.10–3.13, and package build checks without deploying AWS.
 A pushed `v*` version tag runs the same validation before package publication.
-After PyPI succeeds, the release workflow deploys the tagged source to AWS.
-It publishes an immutable ECR image, updates the existing CloudFormation
-stack, and verifies MCP authentication and integration readiness. It preserves
-infrastructure parameters and rejects unrelated infrastructure changes.
-
-AWS deployment requires repository secrets `AWS_ACCESS_KEY_ID`,
-`AWS_SECRET_ACCESS_KEY`, and `MCP_SMOKE_KEY`. See
-[deploy/aws/CI-CD.md](deploy/aws/CI-CD.md) for permissions and operational limits.
-Infrastructure changes use the separate reviewed operator deployment flow.
-AWS deployment is a downstream release job; it does not publish to PyPI itself.
+From 0.4.1, tag releases do not deploy AWS. Existing immutable-image,
+CloudFormation and smoke-check scripts remain manual operator tools; package
+publication does not update the hosted service. See
+[deploy/aws/CI-CD.md](deploy/aws/CI-CD.md) for the separate operator flow.
 
 ## Releases
 
 Package releases start when a `v*` version tag is pushed. Creating a GitHub
 Release manually is not the trigger. `release-pypi.yml` publishes to PyPI using
 Trusted Publishing (OIDC), creates the GitHub Release from the changelog, and
-then publishes to the MCP Registry. AWS deployment runs after successful PyPI
-publication.
+then publishes to the MCP Registry. AWS deployment is a separate manual action.
 
 1. On a feature branch, update `pyproject.toml`,
    `src/deep_agentic_core_mcp/__init__.py`, both version fields in `server.json`,
@@ -98,7 +91,7 @@ publication.
    `git tag -a vX.Y.Z -F <release-notes-file> --cleanup=verbatim`.
 4. Push that tag: `git push origin vX.Y.Z`.
 5. Approve the existing `pypi` environment gate when required, then verify PyPI,
-   GitHub Release, MCP Registry, and AWS deployment jobs.
+   GitHub Release and MCP Registry jobs.
 
-AWS updates through CI only as part of a version-tag release after PyPI succeeds.
-A `main` push does not publish a package or deploy the hosted service.
+Neither a tag release nor a `main` push deploys the hosted service.
+A `main` push does not publish a package.

README.md

@@ -190,7 +190,7 @@ The exported report is MCP-local, not an AIOS artifact. No sibling release or
 dependency floor changes. See the [unified workflows guide](docs/unified-workflows.md)
 and [`examples/unified_release_check.py`](examples/unified_release_check.py).
 
-## MCP Surface (current, `0.4.0`)
+## MCP Surface (current, `0.4.1`)
 
 - `core.health` — rich diagnostics: adapter availability/version, loaded
   tool/resource/prompt counts, workspace root, recent successful calls
@@ -369,11 +369,11 @@ PyPI `0.2.0` release provides local stdio.
 |---|---|---|
 | `.github/workflows/ci.yml` | Pull request | Tests Python 3.10–3.13 and builds distributions; no AWS deployment |
 | `.github/workflows/ci.yml` | Push/merge to `main`, or manual run | Runs tests and package checks; no AWS deployment |
-| `.github/workflows/release-pypi.yml` | Push a `v*` version tag | Runs checks, publishes to PyPI, creates the GitHub Release, publishes MCP Registry metadata, and deploys AWS after PyPI succeeds |
+| `.github/workflows/release-pypi.yml` | Push a `v*` version tag | Runs checks, publishes to PyPI, creates the GitHub Release, publishes MCP Registry metadata; no AWS deployment |
 
-A push to `main` does not publish to PyPI or deploy AWS. A version-tag release
-deploys the same tagged source to AWS only after PyPI publication succeeds. Package releases use PyPI Trusted Publishing;
-AWS deployment uses the dedicated IAM user's GitHub secrets. See
+A push to `main` does not publish to PyPI or deploy AWS. Version-tag releases
+publish packages using PyPI Trusted Publishing. AWS deployment is an explicit
+manual operator action from 0.4.1. See
 [CONTRIBUTING.md](CONTRIBUTING.md) for releases and
 [the AWS CI guide](deploy/aws/CI-CD.md) for deployment setup and limits.
 
@@ -642,3 +642,9 @@ the existing registry discovers faults. Workflows, experiments and fuzzing
 are Python/CLI orchestration, not new MCP tools. Hosted HTTP continues to
 reject `chaos.run_experiment`, so no experiment target is executed by the
 hosted service.
+
+## Sidecar 0.9.0 interoperability
+
+Task-authority/delegation and reader-safe summaries are available in Sidecar.
+See [compatibility and ownership boundaries](docs/sidecar-0.9.md). Existing
+native decisions/correlation stay compatible; no new normative schema is claimed.

ROADMAP.md

@@ -10,8 +10,8 @@ entry below have been updated accordingly; see the audit for full detail.
 
 ## Release Status
 
-Release version: `0.4.0` — see [CHANGELOG.md](CHANGELOG.md). It completes
-Phase 4: `core.compare_runs`, a fail-closed `core.release_check`
+Release version: `0.4.1` (2026-10-19) — see [CHANGELOG.md](CHANGELOG.md). It adds Sidecar 0.9 discovery and manual AWS deployment.
+The 0.4.0 release completed Phase 4: `core.compare_runs`, a fail-closed `core.release_check`
 over Lens, Evals and Chaos evidence, `core.incident_summary` and
 `core.export_report` ([guide](docs/unified-workflows.md)). Hosted image pins and
 sibling dependency ranges are unchanged.
@@ -817,3 +817,9 @@ the existing registry discovers faults. Workflows, experiments and fuzzing
 are Python/CLI orchestration, not new MCP tools. Hosted HTTP continues to
 reject `chaos.run_experiment`, so no experiment target is executed by the
 hosted service.
+
+## Sidecar 0.9.0 interoperability
+
+Task-authority/delegation and reader-safe summaries are available in Sidecar.
+See [compatibility and ownership boundaries](docs/sidecar-0.9.md). Existing
+native decisions/correlation stay compatible; no new normative schema is claimed.

deploy/aws/CI-CD.md

@@ -1,47 +1,16 @@
-# AWS continuous deployment
-
-CI checks pull requests, pushes to `main`, and manual runs without deploying.
-A pushed `v*` version tag starts `release-pypi.yml`, which runs the full test
-matrix and package checks before publishing to PyPI. AWS deployment depends on
-successful PyPI publication and builds the same tagged source. It publishes an
-immutable ECR image, updates ECS through CloudFormation, and verifies the public
-HTTPS endpoint and authenticated MCP integrations. A failed PyPI publication
-prevents AWS deployment.
-
-## GitHub secrets
-
-Configure repository secrets in `DeepAgentLabs/mcp-server`:
-
-- `AWS_ACCESS_KEY_ID` and `AWS_SECRET_ACCESS_KEY`: credentials for the dedicated
-  `deep-agentic-core-mcp-github-deploy` IAM user.
-- `MCP_SMOKE_KEY`: a dedicated application API key for authenticated health checks.
-
-These secrets have been installed using the operator's AWS profile. Region
-`us-east-2` is configured in the workflow. Never put credentials in the repo.
-
-The scoped IAM policy is managed in AWS IAM. It allows ECR publishing,
-CloudFormation change sets on the existing deployment, ECS revision/service
-updates, and passing the two existing runtime roles. Read permissions for load-balancer and CloudFront attributes and deployment
-events allow CloudFormation to resolve existing resource references and report
-failures. Task-definition registration, reads, and deregistration use wildcard
-resources where AWS requires them; service and pass-role permissions are scoped.
-Infrastructure provisioning and database access are handled separately using the operator profile.
-
-## Deployment behavior
-
-Deployments are serialized. Each image tag includes the release tag, run ID,
-and attempt; ECS receives the image digest. Release tags should advance in order;
-rerunning an older release deployment would deploy that older version.
-
-The deployment script uses the existing CloudFormation template and preserves
-all parameters except `ImageUri`, including the domain and cache credentials.
-It refuses changes outside the ECS task definition and service, and refuses
-service replacement. ECS's deployment circuit breaker rolls back tasks that fail
-to become healthy. A failure of the final MCP smoke check fails the workflow;
-it does not automatically roll back an otherwise completed deployment.
-
-For infrastructure changes, use the reviewed operator flow in [README.md](README.md).
-To retry a failed release deployment, rerun its failed job in GitHub Actions.
-To deploy new code, prepare and publish a new version tag. Rotate the IAM access key by updating both repository secrets, then
-remove the old key after a successful run. Rotate or revoke the MCP smoke key
-through its application account and update `MCP_SMOKE_KEY` when rotating.
+# AWS deployment and package releases
+
+From **0.4.1**, GitHub Actions does not deploy AWS. Pull requests and main pushes
+run checks; a `v*` tag validates and publishes PyPI, GitHub Release and MCP Registry
+metadata. Publishing or retrying a release does not update the running AWS service.
+
+The existing Docker build, ECR/CloudFormation deployment and smoke-check scripts
+remain manual operator tools. Follow the reviewed operator flow in
+[README.md](README.md) for an explicit deployment. The service template and runtime
+resources are unchanged by this pipeline change. Do not infer hosted deployment
+version or health from a successful package release.
+
+Standalone image recipes pin Sidecar 0.9.0 for package discovery; they expose no
+live Sidecar task controls. Use an immutable image reference for manual deployments.
+Existing GitHub AWS secrets are not consumed by the release workflow. Credential
+rotation and infrastructure provisioning remain separate operator responsibilities.

deploy/docker/Dockerfile.aws

@@ -8,7 +8,7 @@ COPY src ./src
 # Evals 0.9.0 (the producer of the native report/gate schemas bundled for evals.validate_artifact).
 # Chaos is installed for its exporter and fault catalog only; HTTP always
 # rejects chaos.run_experiment, so no target script is ever executed here.
-RUN pip install --no-cache-dir ".[http]" "agenticlens==0.10.0" "agentic-chaos==0.8.0" "agentic-evals==0.9.0"
+RUN pip install --no-cache-dir ".[http]" "agenticlens==0.10.0" "agentic-chaos==0.8.0" "agentic-evals==0.9.0" "agentic-sidecar==0.9.0"
 COPY scripts/fetch_schema_assets.py /tmp/fetch_schema_assets.py
 RUN python /tmp/fetch_schema_assets.py /opt/workspace/ai-operations-spec
 RUN useradd --create-home --uid 10001 mcp

deploy/docker/Dockerfile.render

@@ -8,7 +8,7 @@ COPY src ./src
 # Evals 0.9.0 (the producer of the native report/gate schemas bundled for evals.validate_artifact).
 # Chaos is installed for its exporter and fault catalog only; HTTP always
 # rejects chaos.run_experiment, so no target script is ever executed here.
-RUN pip install --no-cache-dir ".[http]" "agenticlens==0.10.0" "agentic-chaos==0.8.0" "agentic-evals==0.9.0"
+RUN pip install --no-cache-dir ".[http]" "agenticlens==0.10.0" "agentic-chaos==0.8.0" "agentic-evals==0.9.0" "agentic-sidecar==0.9.0"
 COPY scripts/fetch_schema_assets.py /tmp/fetch_schema_assets.py
 RUN python /tmp/fetch_schema_assets.py /opt/workspace/ai-operations-spec
 RUN useradd --create-home --uid 10001 mcp

deploy/docker/README.md

@@ -28,3 +28,12 @@ docker build -f deploy/docker/Dockerfile.render -t deep-agentic-core-mcp:render
 
 The adjacent `Dockerfile.*.dockerignore` files filter those direct build
 contexts. The helper is the portable option across Docker builders.
+
+
+## Release pipeline from 0.4.1
+
+Tag releases validate, publish PyPI/GitHub Release and publish the MCP Registry.
+Automatic AWS deployment is removed from GitHub Actions. Existing AWS build and
+operator deployment scripts remain available for an explicit manual deployment;
+publishing a package does not update the running AWS service. Standalone image
+recipes include Sidecar 0.9.0 for package discovery, with no live runtime bridge.

docs/index.md

@@ -2,7 +2,7 @@
 
 ## Unified MCP interface for agentic workflows
 
-**Version:** 0.4.0
+**Version:** 0.4.1
 **Maturity:** Alpha
 
 `deep-agentic-core-mcp` is the shared MCP server layer for the DeepAgentLabs ecosystem. It provides one MCP interface for workflow analysis, resilience testing, server diagnostics, and supervision-readiness discovery.

docs/roadmap.md

@@ -5,7 +5,7 @@ interface across the DeepAgentLabs ecosystem.
 
 ## Current Release
 
-**Version:** `0.4.0`
+**Version:** `0.4.1`
 **Released:** 2026-10-19
 
 ## Development Status

docs/sidecar-0.9.md

@@ -0,0 +1,23 @@
+# Sidecar 0.9.0 compatibility
+
+Sidecar 0.9.0 adds explicit task-authority tool bindings, narrowed delegation,
+ancestor revocation, dispatch-time expiry checks, context reader permissions and
+redacted evidence summaries. Private scope goals/values use explicit host delivery.
+
+MCP discovery detects installed task-authority/revocation and redacted-evidence APIs without constructing a task. Discovery does not connect to or govern a runtime. Protected MCP servers/credential issuers remain responsible for remote access revocation; Sidecar registries are process-local.
+
+The core package remains standalone. No new mandatory sibling dependency or
+shared schema is introduced. Constraint-only guardians retain expiry WARN;
+authority-bearing tasks BLOCK on expiry/revocation. Stop revokes descendants;
+restart does not revive grants. Operator intent modification cannot reset a
+delegated grant. Reader/actor identities must come from the trusted harness.
+
+The fixed comparison has 24 offline synthetic runs (12 per variant): Agent Alone
+dispatches 6/6 unsafe proposals; Sidecar dispatches 0/6, with 2/6 false interventions
+on clean gate/provider-outage controls. Zero live inference cost/tokens are
+synthetic accounting, not an economic or production-reliability claim. Raw
+arguments/journals/operator views remain privileged; redaction is allowlisted,
+not arbitrary secret detection or process isolation.
+
+See the [Sidecar v0.9 guide](https://github.com/DeepAgentLabs/agentic-sidecar/blob/v0.9.0/website/docs/v0.9.md)
+and [paired evidence](https://github.com/DeepAgentLabs/agentic-sidecar/blob/v0.9.0/tests/validation/v0_9_supervision.json).

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "deep-agentic-core-mcp"
-version = "0.4.0"
+version = "0.4.1"
 description = "Unified MCP server for AgenticLens and Agentic Chaos workflows."
 readme = "README.md"
 requires-python = ">=3.10"

server.json

@@ -3,7 +3,7 @@
   "name": "io.github.DeepAgentLabs/deep-agentic-core-mcp",
   "title": "Deep Agentic Core MCP",
   "description": "Unified MCP server for AgenticLens and Agentic Chaos capabilities.",
-  "version": "0.4.0",
+  "version": "0.4.1",
   "repository": {
     "url": "https://github.com/DeepAgentLabs/mcp-server",
     "source": "github"
@@ -12,7 +12,7 @@
     {
       "registryType": "pypi",
       "identifier": "deep-agentic-core-mcp",
-      "version": "0.4.0",
+      "version": "0.4.1",
       "transport": {
         "type": "stdio"
       }

src/deep_agentic_core_mcp/__init__.py

@@ -2,4 +2,4 @@
 
 __all__ = ["__version__"]
 
-__version__ = "0.4.0"
+__version__ = "0.4.1"