docs(wiki): generate wiki from docs/ (pages, sidebar, cross-links)

2026-06-26 12:58:15 +02:00
parent 10883754b7
commit 1b0d4b2957
18 changed files with 1082 additions and 1 deletions
+118
@@ -0,0 +1,118 @@
# API Reference
## Core Endpoints
- `GET /`: server metadata.
- `GET /health`: health probe.
- `GET /metrics`: Prometheus metrics (when enabled).
## OAuth Discovery and Token Exchange
- `GET /.well-known/oauth-protected-resource`
- Returns OAuth protected resource metadata used by MCP clients.
- `GET /.well-known/oauth-authorization-server`
- Returns OAuth authorization server metadata.
- `POST /register`
- Registers an OAuth client and persists the client metadata.
- `POST /oauth/token`
- Proxies OAuth authorization-code token exchange to Gitea.
## MCP Endpoints
- `GET /mcp/tools`: list tool definitions.
- `GET /mcp` and `POST /mcp`: streamable HTTP transport.
- `GET /mcp/sse` and `POST /mcp/sse`: MCP SSE transport alias.
- `POST /mcp/tool/call`: direct tool-call endpoint.
Authentication requirements:
- MCP tool execution requires `Authorization: Bearer <token>`.
- Missing or invalid tokens return `401` with:
- `WWW-Authenticate: Bearer resource_metadata="<absolute metadata url>", scope="read:repository"`
Scope requirements:
- Read tools require `read:repository`.
- Write tools require `write:repository`.
- Insufficient scope returns `403`.
## Automation Endpoints
- `POST /automation/webhook`: ingest policy-controlled webhook events.
- `POST /automation/jobs/run`: run policy-controlled automation jobs.
## Read Tools
- `list_repositories`
- `get_repository_info` (`owner`, `repo`)
- `get_file_tree` (`owner`, `repo`, optional `ref`, `recursive`)
- `get_file_contents` (`owner`, `repo`, `filepath`, optional `ref`)
- `search_code` (`owner`, `repo`, `query`, optional `ref`, `page`, `limit`)
- `list_commits` (`owner`, `repo`, optional `ref`, `page`, `limit`)
- `get_commit_diff` (`owner`, `repo`, `sha`)
- `compare_refs` (`owner`, `repo`, `base`, `head`)
- `list_issues` (`owner`, `repo`, optional `state`, `page`, `limit`, `labels`)
- `get_issue` (`owner`, `repo`, `issue_number`)
- `list_pull_requests` (`owner`, `repo`, optional `state`, `page`, `limit`)
- `get_pull_request` (`owner`, `repo`, `pull_number`)
- `list_labels` (`owner`, `repo`, optional `page`, `limit`)
- `list_tags` (`owner`, `repo`, optional `page`, `limit`)
- `list_releases` (`owner`, `repo`, optional `page`, `limit`)
- `list_pull_request_files` (`owner`, `repo`, `pull_number`, optional `page`, `limit`)
- `list_pull_request_commits` (`owner`, `repo`, `pull_number`, optional `page`, `limit`)
- `list_issue_comments` (`owner`, `repo`, `issue_number`, optional `page`, `limit`)
- `list_branches` (`owner`, `repo`, optional `page`, `limit`)
- `get_branch` (`owner`, `repo`, `branch`)
- `get_release` (`owner`, `repo`, `release_id`)
- `get_latest_release` (`owner`, `repo`)
- `list_milestones` (`owner`, `repo`, optional `state`, `page`, `limit`)
- `get_commit_status` (`owner`, `repo`, `sha`)
- `list_org_repositories` (`org`, optional `page`, `limit`)
- `list_organizations` (optional `page`, `limit`)
- `get_repo_languages` (`owner`, `repo`)
- `list_repo_topics` (`owner`, `repo`)
## Write Tools (Write Mode Required)
- `create_issue` (`owner`, `repo`, `title`, optional `body`, `labels`, `assignees`, `milestone`)
- `update_issue` (`owner`, `repo`, `issue_number`, one or more of `title`, `body`, `state`, `milestone`)
- `create_issue_comment` (`owner`, `repo`, `issue_number`, `body`)
- `create_pr_comment` (`owner`, `repo`, `pull_number`, `body`)
- `add_labels` (`owner`, `repo`, `issue_number`, `labels` by name)
- `remove_labels` (`owner`, `repo`, `issue_number`, `labels` by name)
- `assign_issue` (`owner`, `repo`, `issue_number`, `assignees`)
- `create_label` (`owner`, `repo`, `name`, `color` hex e.g. `#00aabb`, optional `description`, `exclusive`)
- `update_label` (`owner`, `repo`, `name`, one or more of `new_name`, `color`, `description`)
- `create_pull_request` (`owner`, `repo`, `title`, `head`, `base`, optional `body`)
- `create_release` (`owner`, `repo`, `tag_name`, optional `name`, `body`, `draft`, `prerelease`, `target`)
- `edit_release` (`owner`, `repo`, `release_id`, one or more of `name`, `body`, `draft`, `prerelease`)
- `create_branch` (`owner`, `repo`, `new_branch_name`, optional `old_branch_name`)
- `create_milestone` (`owner`, `repo`, `title`, optional `description`, `due_on`)
- `edit_issue_comment` (`owner`, `repo`, `comment_id`, `body`)
Not supported by design: merge, branch/label/release deletion, force push, repo/admin
management.
Note: `create_issue`, `add_labels`, and `remove_labels` accept label **names**; the
server resolves them to Gitea label ids and returns a clear error for unknown labels.
Note: the `milestone` argument on `create_issue`/`update_issue` accepts either a numeric
milestone **id** or a milestone **title** (resolved case-insensitively against open and
closed milestones; unknown titles return a clear error). On `update_issue`, `milestone: 0`
clears the issue's milestone. Gitea Projects (Kanban boards) are intentionally unsupported:
the Gitea REST API exposes no project endpoints.
## Validation and Limits
- All tool argument schemas reject unknown fields.
- List responses are capped by `MAX_TOOL_RESPONSE_ITEMS`.
- Text payloads are capped by `MAX_TOOL_RESPONSE_CHARS`.
- File reads are capped by `MAX_FILE_SIZE_BYTES`.
## Error Model
- Auth error: HTTP `401`.
- Policy/scope denial: HTTP `403`.
- Validation error: HTTP `400`.
- Rate limit: HTTP `429`.
- Internal errors: HTTP `500` (no stack traces in production).
+175
@@ -0,0 +1,175 @@
# Architecture
## Overview
AegisGitea MCP is a Python 3.10+ application built on **FastAPI**. It acts as a bridge between an AI client (such as Claude, Claude Code, or Cowork) and a self-hosted Gitea instance, implementing the [Model Context Protocol (MCP)](https://modelcontextprotocol.io).
```
AI Client (Claude / Claude Code / Cowork)
│ HTTP (Authorization: Bearer <key>)
┌────────────────────────────────────────────┐
│ FastAPI Server │
│ server.py │
│ - Route: GET/POST /mcp │
│ - Route: POST /mcp/tool/call │
│ - Route: GET /mcp/tools │
│ - Route: GET /health │
│ - Streamable HTTP transport │
│ - Legacy SSE alias (GET/POST /mcp/sse) │
└───────┬───────────────────┬────────────────┘
│ │
┌────▼────┐ ┌────▼──────────────┐
│ auth │ │ Tool dispatcher │
│ auth.py│ │ (server.py) │
└────┬────┘ └────────┬──────────┘
│ │
│ ┌────────▼──────────┐
│ │ Tool handlers │
│ │ tools/repo.py │
│ └────────┬──────────┘
│ │
│ ┌────────▼──────────┐
│ │ GiteaClient │
│ │ gitea_client.py │
│ └────────┬──────────┘
│ │ HTTPS
│ ▼
│ Gitea instance
┌────▼────────────────────┐
│ AuditLogger │
│ audit.py │
│ /var/log/aegis-mcp/ │
│ audit.log │
└─────────────────────────┘
```
---
## Source Modules
### `server.py`
The entry point and FastAPI application. Responsibilities:
- Defines all HTTP routes
- Reads configuration on startup and initialises `GiteaClient`
- Applies authentication middleware to protected routes
- Dispatches tool calls to the appropriate handler function
- Handles CORS
### `auth.py`
API key validation. Responsibilities:
- `APIKeyValidator` class: holds the set of valid keys, tracks failed attempts per IP
- Constant-time comparison to prevent timing side-channels
- Rate limiting: blocks IPs that exceed `MAX_AUTH_FAILURES` within `AUTH_FAILURE_WINDOW`
- Helper functions for key generation and hashing
- Singleton pattern (`get_validator()`) with test-friendly reset (`reset_validator()`)
### `config.py`
Pydantic `BaseSettings` model. Responsibilities:
- Loads all configuration from environment variables or `.env`
- Validates values (log level enum, token format, key minimum length)
- Parses comma-separated `MCP_API_KEYS` into a list
- Exposes computed properties (e.g. base URL for Gitea API)
### `gitea_client.py`
Async HTTP client for the Gitea API. Responsibilities:
- Wraps `httpx.AsyncClient` with bearer token authentication
- Maps HTTP status codes to typed exceptions (`GiteaAuthenticationError`, `GiteaNotFoundError`, etc.)
- Enforces file size limit before returning file contents
- Logs all API calls to the audit logger
Key methods:
| Method | Gitea endpoint |
|---|---|
| `get_current_user()` | `GET /api/v1/user` |
| `list_repositories()` | `GET /api/v1/user/repos` |
| `get_repository()` | `GET /api/v1/repos/{owner}/{repo}` |
| `get_file_contents()` | `GET /api/v1/repos/{owner}/{repo}/contents/{path}` |
| `get_tree()` | `GET /api/v1/repos/{owner}/{repo}/git/trees/{ref}` |
### `audit.py`
Structured audit logging using `structlog`. Responsibilities:
- Initialises a `structlog` logger writing JSON to the configured log file
- `log_tool_invocation()`: records tool calls with result and correlation ID
- `log_access_denied()`: records failed authentication
- `log_security_event()`: records rate limit triggers and other security events
- Auto-generates UUID correlation IDs when none is provided
### `mcp_protocol.py`
MCP data models and tool registry. Responsibilities:
- Pydantic models: `MCPTool`, `MCPToolCallRequest`, `MCPToolCallResponse`, `MCPListToolsResponse`
- `AVAILABLE_TOOLS` list: the canonical list of tools exposed to clients
- `get_tool_by_name()`: lookup helper used by the dispatcher
### `tools/repository.py`
Concrete tool handler functions. Responsibilities:
- `list_repositories_tool()`: calls `GiteaClient.list_repositories()`, formats the result
- `get_repository_info_tool()`: calls `GiteaClient.get_repository()`, formats metadata
- `get_file_tree_tool()`: calls `GiteaClient.get_tree()`, flattens to a list of paths
- `get_file_contents_tool()`: calls `GiteaClient.get_file_contents()`, decodes base64
All handlers return a plain string. `server.py` wraps this in an `MCPToolCallResponse`.
---
## Request Lifecycle
```
1. Client sends POST /mcp/tool/call
2. FastAPI routes the request to the tool-call handler in server.py
3. OAuth middleware validates the Bearer token via Gitea OIDC/JWKS or userinfo
├── Fail → AuditLogger.log_access_denied() → HTTP 401 / 429
└── Pass → continue
4. AuditLogger.log_tool_invocation(status="pending")
5. Tool dispatcher looks up the tool by name (mcp_protocol.get_tool_by_name)
6. Policy engine checks read/write mode and repository/path policy
7. If GITEA_TOKEN is configured, service-PAT authz checks
GET /repos/{owner}/{repo}/collaborators/{user}/permission
8. Tool handler function (tools/repository.py) is called
9. GiteaClient makes an async HTTP call to the Gitea API
10. Result (or error) is returned to server.py
11. AuditLogger.log_tool_invocation(status="success" | "error")
12. MCPToolCallResponse is returned to the client
```
---
## Key Design Decisions
**Read by default, writes opt-in.** Read tools are available by default. Write-capable tools require `WRITE_MODE=true`, repository write policy/whitelist approval, and `write:repository` authorization.
**Gitea controls repository access.** Without `GITEA_TOKEN`, Gitea enforces repository permissions on API calls made with the user's token. With `GITEA_TOKEN`, the service PAT can only execute after the server verifies the requesting user's actual repository permission through Gitea and writes an audit denial if the check fails.
**Public tool discovery.** `GET /mcp/tools` requires no authentication so that MCP clients can discover the available tools without credentials. All other endpoints require authentication.
**Minimal persisted state.** The audit log is persisted for tamper-evident review. Dynamic OAuth client registrations are persisted when DCR is enabled. Rate limit counters and short-lived authz caches are in-memory and reset on restart.
**Async throughout.** FastAPI + `httpx.AsyncClient` means all Gitea API calls are non-blocking, allowing the server to handle concurrent requests efficiently.
+53
@@ -0,0 +1,53 @@
# Audit Logging
## Design
Audit logs are append-only JSON lines with hash chaining:
- `prev_hash`: previous entry hash.
- `entry_hash`: hash of current entry payload + previous hash.
This makes tampering detectable.
## Event Types
- `tool_invocation`
- `access_denied`
- `security_event`
Each event includes timestamps and correlation context.
## Integrity Validation
Use:
```bash
python3 scripts/validate_audit_log.py --path /var/log/aegis-mcp/audit.log
```
Exit code `0` indicates valid chain, non-zero indicates tamper/corruption.
## Operational Expectations
- Persist audit logs to durable storage.
- Protect write permissions (service account only).
- Validate integrity during incident response and release checks.
## Rotation
The server appends to a single audit file and does not rotate it in process — rotating
mid-stream would break the `prev_hash`/`entry_hash` chain. Manage growth externally with
`logrotate` using `copytruncate` so the open file handle keeps appending:
```
/var/log/aegis-mcp/audit.log {
weekly
rotate 12
compress
missingok
notifempty
copytruncate
}
```
Run `scripts/validate_audit_log.py` against each rotated segment to confirm the chain
remains intact across rotations before archiving.
+27
@@ -0,0 +1,27 @@
# Automation
## Scope
Current automation capabilities:
- Webhook ingestion endpoint (`POST /automation/webhook`).
- On-demand scheduled-job execution endpoint (`POST /automation/jobs/run`).
- Dependency hygiene scan job scaffold (`dependency_hygiene_scan`).
- Stale issue detection job (`stale_issue_detection`).
- Auto issue creation job scaffold (`auto_issue_creation`, write-mode and policy required).
Planned extensions:
- Background scheduler orchestration.
## Control Requirements
All automation must be:
- Policy-controlled.
- Independently disableable.
- Fully audited.
- Explicitly documented with runbook guidance.
## Enablement
- `AUTOMATION_ENABLED=true` to allow automation endpoints.
- `AUTOMATION_SCHEDULER_ENABLED=true` reserved for future built-in scheduler loop.
- Policy rules must allow automation pseudo-tools (`automation_*`) per repository.
+72
@@ -0,0 +1,72 @@
# Configuration
Copy `.env.example` to `.env` and set values before starting:
```bash
cp .env.example .env
```
## OAuth/OIDC Settings (Primary)
| Variable | Required | Default | Description |
|---|---|---|---|
| `GITEA_URL` | Yes | - | Base URL of your Gitea instance |
| `OAUTH_MODE` | No | `false` | Enables OAuth-oriented validation settings |
| `GITEA_OAUTH_CLIENT_ID` | Yes when `OAUTH_MODE=true` | - | OAuth client id |
| `GITEA_OAUTH_CLIENT_SECRET` | Yes when `OAUTH_MODE=true` | - | OAuth client secret |
| `OAUTH_EXPECTED_AUDIENCE` | No | empty | Additional accepted JWT audience beyond the MCP resource and Gitea client id |
| `OAUTH_CACHE_TTL_SECONDS` | No | `300` | OIDC discovery/JWKS cache TTL |
| `OAUTH_STATE_SECRET` | Yes when `OAUTH_MODE=true` | - | HMAC secret for signed OAuth state wrappers; must be at least 32 characters (e.g. `openssl rand -hex 32`) |
| `OAUTH_REDIRECT_ALLOWLIST` | No | empty | Additional allowed redirect URIs for OAuth clients |
## MCP Server Settings
| Variable | Required | Default | Description |
|---|---|---|---|
| `MCP_HOST` | No | `127.0.0.1` | Interface to bind to |
| `MCP_PORT` | No | `8080` | Port to listen on |
| `PUBLIC_BASE_URL` | No | empty | Public HTTPS base URL advertised in OAuth metadata (recommended behind reverse proxy) |
| `ALLOW_INSECURE_BIND` | No | `false` | Explicit opt-in required for `0.0.0.0` bind |
| `LOG_LEVEL` | No | `INFO` | `DEBUG`, `INFO`, `WARNING`, `ERROR`, `CRITICAL` |
| `STARTUP_VALIDATE_GITEA` | No | `true` | Validate OIDC discovery endpoint at startup |
| `DCR_ENABLED` | No | `true` | Enable dynamic client registration at `/register` |
| `DCR_STORAGE_PATH` | No | `/var/lib/aegis-mcp/dcr_clients.json` | Persisted OAuth client registry path. Written with owner-only (`0o600`) permissions on POSIX hosts |
## Security and Limits
| Variable | Required | Default | Description |
|---|---|---|---|
| `MAX_AUTH_FAILURES` | No | `5` | Failed auth attempts before rate limiting |
| `AUTH_FAILURE_WINDOW` | No | `300` | Window in seconds for auth failure counting |
| `RATE_LIMIT_PER_MINUTE` | No | `60` | Per-IP request limit |
| `TOKEN_RATE_LIMIT_PER_MINUTE` | No | `120` | Per-token request limit |
| `MAX_FILE_SIZE_BYTES` | No | `1048576` | Max file payload returned by read tools |
| `MAX_TOOL_RESPONSE_ITEMS` | No | `200` | Max list items in tool responses |
| `MAX_TOOL_RESPONSE_CHARS` | No | `20000` | Max chars in text fields |
| `REQUEST_TIMEOUT_SECONDS` | No | `30` | Upstream timeout for Gitea calls |
| `SECRET_DETECTION_MODE` | No | `mask` | `off`, `mask`, `block` |
| `REPO_AUTHZ_CACHE_TTL_SECONDS` | No | `60` | TTL for cached per-user repository permission checks |
## Write Mode
| Variable | Required | Default | Description |
|---|---|---|---|
| `WRITE_MODE` | No | `false` | Enables write tools |
| `WRITE_REPOSITORY_WHITELIST` | Required if write mode enabled and allow-all disabled | empty | Comma-separated `owner/repo` allow list |
| `WRITE_ALLOW_ALL_TOKEN_REPOS` | No | `false` | Allow all repos accessible by token |
## Automation
| Variable | Required | Default | Description |
|---|---|---|---|
| `AUTOMATION_ENABLED` | No | `false` | Enables automation endpoints |
| `AUTOMATION_SCHEDULER_ENABLED` | No | `false` | Enables scheduler loop |
| `AUTOMATION_STALE_DAYS` | No | `30` | Age threshold for stale issue checks |
## Legacy Compatibility Variables
These are retained for compatibility but not used for OAuth-protected MCP tool execution:
- `GITEA_TOKEN`
- `MCP_API_KEYS`
- `AUTH_ENABLED`
+57
@@ -0,0 +1,57 @@
# Deployment
## Secure Defaults
- Default bind is `127.0.0.1`.
- Binding `0.0.0.0` requires `ALLOW_INSECURE_BIND=true`.
- Write mode disabled by default.
- Policy checks run before tool execution.
- OAuth-protected MCP challenge responses are enabled by default for tool calls.
## Local Development
```bash
make install-dev
cp .env.example .env
make run
```
## Docker
Use `docker/Dockerfile`:
- Multi-stage image build.
- Non-root runtime user.
- Production env flags (`NODE_ENV=production`, `ENVIRONMENT=production`).
- Only required app files copied.
- Healthcheck on `/health`.
Run examples:
```bash
docker compose --profile prod up -d
docker compose --profile dev up -d
```
## CI/CD (Gitea Workflows)
Workflows live in `.gitea/workflows/`:
- `lint.yml`: ruff + format checks + mypy.
- `test.yml`: lint + tests + coverage fail-under `80`.
- `docker.yml`: lint + test + docker smoke-test gating; image publish on push to `main`/`dev` and on approved PR review targeting `main`/`dev`; tags include commit SHA plus `latest` (`main`) or `dev` (`dev`).
Docker publish settings:
- `vars.PUSH_IMAGE=true` enables registry push.
- `vars.REGISTRY_IMAGE` sets the target image name (for example `registry.example.com/org/aegis-gitea-mcp`).
- `vars.REGISTRY_HOST` is optional and overrides the login host detection.
- `secrets.REGISTRY_USER` and `secrets.REGISTRY_TOKEN` are required when push is enabled.
## Production Recommendations
- Place MCP behind TLS reverse proxy.
- Set `PUBLIC_BASE_URL=https://<your-mcp-domain>` so OAuth metadata advertises HTTPS endpoints.
- Restrict inbound traffic to expected clients.
- Persist and monitor audit logs.
- Monitor `/metrics` and auth-failure events.
- Rotate OAuth client credentials when required.
+130
@@ -0,0 +1,130 @@
# Getting Started
## Prerequisites
- Python 3.10 or higher
- A running Gitea instance
- A Gitea OAuth2 application for this MCP server
- `make` (optional but recommended)
## 1. Install
```bash
git clone <repo-url>
cd AegisGitea-MCP
# Install production dependencies
make install
# Or install with dev dependencies (for testing and linting)
make install-dev
```
To install manually without `make`:
```bash
python -m venv venv
source venv/bin/activate # Linux/macOS
# or: venv\Scripts\activate # Windows
pip install -e .
# dev: pip install -e ".[dev]"
```
## 2. Create a Gitea OAuth2 Application
1. In Gitea, open **User Settings > Applications**.
2. Create an OAuth2 application for AegisGitea-MCP.
3. Set the redirect URI to `https://<host>/oauth/callback`.
4. Copy the client ID and client secret.
## 3. Configure
Copy the example environment file and fill in your values:
```bash
cp .env.example .env
```
Minimum OAuth settings in `.env`:
```env
GITEA_URL=https://gitea.example.com
OAUTH_MODE=true
GITEA_OAUTH_CLIENT_ID=<your-gitea-oauth-client-id>
GITEA_OAUTH_CLIENT_SECRET=<your-gitea-oauth-client-secret>
PUBLIC_BASE_URL=https://<host>
OAUTH_STATE_SECRET=<random-32-byte-minimum-secret>
```
`GITEA_TOKEN` is optional. If it is set, use a narrowly scoped service PAT and only grant it repository access you are prepared to expose after per-user authorization checks. If it is not set, Gitea REST calls use the authenticated user's OAuth token directly.
See [Configuration](Configuration) for the full list of settings.
## 4. Optional Standard API Key Mode
For non-OAuth deployments, configure `GITEA_TOKEN` and `MCP_API_KEYS`. Generate an API key with:
```bash
make generate-key
# or: python scripts/generate_api_key.py
```
Copy the printed key into `MCP_API_KEYS` in your `.env` file and set `OAUTH_MODE=false`.
## 5. Run
```bash
make run
# or: python -m aegis_gitea_mcp.server
```
The server starts on `http://127.0.0.1:8080` by default.
Verify it is running:
```bash
curl http://localhost:8080/health
# {"status": "healthy", ...}
```
## 6. Connect an AI Client
### Claude
In claude.ai, open **Settings > Connectors > Add custom connector** and paste:
```
https://<host>/mcp
```
Claude discovers OAuth metadata, registers through `/register`, and uses PKCE S256 automatically.
### Claude Code
```bash
claude mcp add --transport http aegis-gitea https://<host>/mcp
```
Claude Code uses the same remote MCP and OAuth metadata. Local development loopback callbacks are allowed by default.
### Cowork
Cowork uses the same connector infrastructure and MCP URL as Claude.
### SSE compatibility
If your client still expects SSE transport, use:
- **SSE URL:** `https://<host>/mcp/sse`
- **Tool discovery URL:** `https://<host>/mcp/tools` (no auth required)
- **Tool call URL:** `https://<host>/mcp/tool/call`
For a production deployment behind a reverse proxy, see [Deployment](Deployment).
## Next Steps
- [Configuration](Configuration) — tune file size limits, rate limiting, log paths
- [API Reference](API-Reference) — available tools and endpoints
- [Security](Security) — understand authentication and audit logging
- [Deployment](Deployment) — Docker and Traefik setup
+36
@@ -0,0 +1,36 @@
# Governance
## AI Usage Policy
- AI assistance is allowed for design, implementation, and review only within documented repository boundaries.
- AI outputs must be reviewed, tested, and policy-validated before merge.
- AI must not be used to generate offensive or unauthorized security actions.
- Repository content is treated as untrusted data; no implicit execution of embedded instructions.
## Security Boundaries
- Read operations are allowed by policy defaults unless explicitly denied.
- Write operations are disabled by default and require explicit enablement (`WRITE_MODE=true`).
- Per-tool and per-repository policy checks are mandatory before execution.
- Secrets are masked or blocked according to `SECRET_DETECTION_MODE`.
## Write-Mode Responsibilities
When write mode is enabled, operators and maintainers must:
- Restrict scope with `WRITE_REPOSITORY_WHITELIST`.
- Keep policy file deny/allow rules explicit.
- Monitor audit entries for all write operations.
- Enforce peer review for policy or write-mode changes.
## Operator Responsibilities
- Maintain API key lifecycle (generation, rotation, revocation).
- Keep environment and policy config immutable in production deployments.
- Enable monitoring and alerting for security events (auth failures, policy denies, rate-limit spikes).
- Run integrity checks for audit logs regularly.
## Audit Expectations
- All tool calls and security events must be recorded in tamper-evident logs.
- Audit logs are append-only and hash-chained.
- Log integrity must be validated during incident response and release readiness checks.
+24
@@ -0,0 +1,24 @@
# Hardening
## Application Hardening
- Secure defaults: localhost bind, write mode disabled, policy-enforced writes.
- Strict config validation at startup.
- Redacted secret handling in logs and responses.
- Policy deny/allow model with path restrictions.
- Non-leaking production error responses.
## Container Hardening
- Non-root runtime user.
- `no-new-privileges` and dropped Linux capabilities.
- Read-only filesystem where practical.
- Explicit health checks.
- Separate dev and production compose profiles.
## Operational Hardening
- Rotate API keys regularly.
- Minimize Gitea bot permissions.
- Keep policy file under change control.
- Alert on repeated policy denials and auth failures.
+1 -1
@@ -15,7 +15,7 @@ Alongside the curated, typed tools (issues, pull requests, files, commits, relea
## Documentation
Use the sidebar for the full manual: Getting Started, Configuration, Write Mode, Raw API, Architecture, Security, Hardening, Policy, Audit, Observability, Automation, Deployment, Governance, Troubleshooting, API Reference and Roadmap.
Use the sidebar for the full manual: Getting Started, Configuration, Write Mode, Architecture, Security, Hardening, Policy, Audit, Observability, Automation, Deployment, Governance, Troubleshooting, API Reference and Roadmap.
---
+48
@@ -0,0 +1,48 @@
# Observability
## Logging
- Structured JSON logs.
- Request correlation via `X-Request-ID`.
- Security events and policy denials are audit logged.
### Structured event helpers
`logging_utils` exposes reusable helpers so endpoints emit consistent,
secret-safe structured events instead of ad-hoc inline logging:
- `log_event(logger, level, event, **context)` — emit a named event with a
`context` mapping; keys in `SENSITIVE_CONTEXT_KEYS` (e.g. `token`,
`authorization`, `password`) are masked as `***`.
- `log_nullable_field(logger, event, field, value)` — record whether a parsed
response field is `None` and its runtime type, without dumping its contents.
- `sanitize_context(context)` — the masking primitive used by the above.
The `context` mapping is serialized into the JSON log payload under a `context`
key. These run at `DEBUG`, so they are silent unless `LOG_LEVEL=DEBUG`.
`get_issue` is instrumented with these helpers (`get_issue.start`,
`get_issue.payload_shape`, `get_issue.field_check`) to make nullable-field
parsing failures diagnosable. The same pattern can be reused for other
parsing-heavy endpoints (`get_pull_request`, `list_issues`, `get_commit_diff`).
## Metrics
Prometheus-compatible endpoint: `GET /metrics`.
Current metrics:
- `aegis_http_requests_total{method,path,status}`
- `aegis_tool_calls_total{tool,status}`
- `aegis_tool_duration_seconds_sum{tool}`
- `aegis_tool_duration_seconds_count{tool}`
## Tracing and Correlation
- Request IDs propagate in response header (`X-Request-ID`).
- Tool-level correlation IDs included in MCP responses.
## Operational Guidance
- Alert on spikes in 401/403/429 rates.
- Alert on repeated `access_denied` and auth-rate-limit events.
- Track tool latency trends for incident triage.
+52
@@ -0,0 +1,52 @@
# Policy Engine
## Overview
Aegis uses a YAML policy engine to authorize tool execution before any Gitea API call is made.
## Behavior Summary
- Global tool allow/deny supported.
- Per-repository tool allow/deny supported.
- Optional repository path allow/deny supported.
- Write operations are denied by default.
- Write operations also require `WRITE_MODE=true` and either:
- `WRITE_REPOSITORY_WHITELIST` match, or
- `WRITE_ALLOW_ALL_TOKEN_REPOS=true`.
## Example Configuration
```yaml
defaults:
read: allow
write: deny
tools:
deny:
- search_code
repositories:
acme/service-a:
tools:
allow:
- get_file_contents
- list_commits
paths:
allow:
- src/*
deny:
- src/secrets/*
```
## Failure Behavior
- Invalid YAML or invalid schema: startup failure (fail closed).
- Denied tool call: HTTP `403` + audit `access_denied` entry.
- Path traversal attempt in path-scoped tools: denied by validation/policy checks.
## Operational Guidance
- Keep policy files version-controlled and code-reviewed.
- Prefer explicit deny entries for sensitive tools.
- Use repository-specific allow lists for high-risk environments.
- Test policy updates in staging before production rollout.
+72
@@ -0,0 +1,72 @@
# Roadmap
## High-Level Evolution Plan
1. Hardened read-only gateway baseline.
2. Policy-driven authorization and observability.
3. Controlled write-mode rollout.
4. Automation and event-driven workflows.
5. Continuous hardening and enterprise controls.
## Threat Model Updates
- Primary threats: credential theft, over-permissioned automation, prompt injection via repo data, policy bypass, audit tampering.
- Secondary threats: denial-of-service, misconfiguration drift, unsafe deployment defaults.
## Security Model
- API key authentication + auth failure throttling.
- Per-IP and per-token request rate limits.
- Secret detection and outbound sanitization.
- Tamper-evident audit logs with integrity verification.
- No production stack-trace disclosure.
## Policy Model
- YAML policy with global and per-repository allow/deny rules.
- Optional path restrictions for file-oriented tools.
- Default write deny.
- Write-mode repository whitelist enforcement.
## Capability Matrix Concept
- `Read` capabilities: enabled by default but policy-filtered.
- `Write` capabilities: disabled by default, policy + whitelist gated.
- `Automation` capabilities: disabled by default, policy-controlled.
## Audit Log Design
- JSON lines.
- `prev_hash` + `entry_hash` chain.
- Correlation/request IDs for traceability.
- Validation script for chain integrity.
## Write-Mode Architecture
- Separate write tool set with strict schemas.
- Global toggle (`WRITE_MODE`) + per-repo whitelist.
- Policy engine still authoritative.
- No merge, branch deletion, or force push endpoints.
## Deployment Architecture
- Non-root container runtime.
- Read-only filesystem where practical.
- Explicit opt-in for insecure bind.
- Separate dev and prod compose profiles.
## Observability Architecture
- Structured JSON logs with request correlation.
- Prometheus-compatible `/metrics` endpoint.
- Tool execution counters and duration aggregates.
## Risk Analysis
- Highest risk: write-mode misuse and policy misconfiguration.
- Mitigations: deny-by-default, whitelist, audit chain, tests, docs, reviews.
## Extensibility Notes
- Add new tools only through schema + policy + docs + tests path.
- Keep transport-agnostic execution core for webhook/scheduler integrations.
+58
@@ -0,0 +1,58 @@
# Security
## Core Controls
- OAuth2/OIDC bearer-token authentication for MCP tool execution.
- OIDC discovery + JWKS validation cache for JWT tokens.
- Userinfo validation fallback for opaque OAuth tokens.
- Scope enforcement:
- `read:repository` for read tools.
- `write:repository` for write tools.
- Policy engine checks before tool execution.
- Per-IP and per-token rate limiting.
- Strict schema validation (`extra=forbid`).
- Tamper-evident audit logging with hash chaining.
- Secret sanitization for logs and tool output.
- Production-safe error responses (no internal stack traces).
## Threat Model
### Why shared bot tokens are dangerous
- A single leaked bot token can expose all repositories that bot can access.
- Access is not naturally bounded per end user.
- Blast radius is large and cross-tenant.
### Why token-in-URL is insecure
- URLs can be captured by reverse proxy logs, browser history, referer headers, and monitoring pipelines.
- Bearer tokens must be passed in `Authorization` headers only.
### Why per-user OAuth reduces lateral access
- Each MCP request executes with the signed-in user token.
- Gitea authorization stays source-of-truth for repository visibility.
- A compromised token is limited to that users permissions.
## Prompt Injection Hardening
Repository content is treated as untrusted data.
- Tool outputs are bounded and sanitized.
- No instructions from repository text are executed.
- Text fields are size-limited before returning to LLM clients.
## Secret Detection
Detected classes include:
- API key and token patterns.
- JWT-like tokens.
- Private key block markers.
- Common provider credential formats.
Behavior:
- `SECRET_DETECTION_MODE=mask`: redact in place.
- `SECRET_DETECTION_MODE=block`: replace secret-bearing values.
- `SECRET_DETECTION_MODE=off`: disable sanitization (not recommended).
+76
@@ -0,0 +1,76 @@
# Troubleshooting
## "Internal server error (-32603)" from Claude
**Symptom:** Claude shows `Internal server error` with JSON-RPC error code `-32603` when trying to use Gitea tools.
**Cause:** In user-token mode, the OAuth token stored by the client may have been issued without Gitea API scopes (e.g. `read:repository`). In service-PAT mode, the call may fail because the authenticated user does not have the required repository permission or the permission probe cannot be completed.
**Fix:**
1. In Gitea: Go to **Settings > Applications > Authorized OAuth2 Applications** and revoke the MCP application.
2. In Claude: disconnect the MCP server and authenticate again.
3. Re-authorize: Use the MCP connector again. It will trigger a fresh OAuth flow. For repository-targeted calls in service-PAT mode, also verify the signed-in Gitea user has read/write access to the target repository.
**Verification:** Check the server logs for `oauth_auth_summary`. A working token shows:
```
oauth_auth_summary: api_probe=pass login=alice
```
A scopeless token shows:
```
oauth_token_lacks_api_scope: status=403 login=alice
```
## "Gitea rejected the API call" (403)
**Symptom:** Tool calls return 403 with a message about re-authorizing.
**Cause:** The OAuth token does not have the required API scope in user-token mode, or the per-user repository permission check denied the request in service-PAT mode.
**Fix:** Revoke and re-authorize if the token lacks API scope. If the error mentions repository permission, grant the signed-in Gitea user the required repository access or use a repository they can access.
## Claude caches stale tokens
**Symptom:** After fixing the OAuth configuration, Claude still sends the old token.
**Cause:** The client caches access tokens and doesn't automatically re-authenticate when the server configuration changes.
**Fix:**
1. Disconnect the server in the client.
2. Start a new conversation and use the integration again - this forces a fresh OAuth flow.
## How OAuth scopes work with Gitea
Gitea's OAuth2/OIDC implementation uses **granular scopes** for API access:
| Scope | Access |
|-------|--------|
| `read:repository` | Read repositories, issues, PRs, files |
| `write:repository` | Create/edit issues, PRs, comments, files |
| `openid` | OIDC identity (login, email) |
When an OAuth application requests authorization, the `scope` parameter in the authorize URL determines what permissions the resulting token has. If only OIDC scopes are requested (e.g. `openid profile email`), the token can establish identity but may not be usable for direct Gitea REST calls. When `GITEA_TOKEN` is configured, the server uses OIDC for identity and checks the user's repository permission before using the service PAT.
The MCP server's OAuth metadata controls which scopes the client requests. Ensure it includes:
```yaml
scopes:
read:repository: "Read access to Gitea repositories"
write:repository: "Write access to Gitea repositories"
```
## Reading the `oauth_auth_summary` log
Every authenticated request emits a structured log line:
| Field | Description |
|-------|-------------|
| `token_type` | `jwt` or `opaque` |
| `scopes_observed` | Scopes extracted from the token/userinfo |
| `scopes_effective` | Final scopes after implicit grants |
| `api_probe` | `pass`, `fail:403`, `fail:401`, `skip:cached`, `skip:error` |
| `login` | Gitea username |
- `api_probe=pass` — token works for Gitea API calls
- `api_probe=fail:403` — token lacks API scopes, request rejected with re-auth guidance
- `api_probe=skip:cached` — previous probe passed, cached result used
- `api_probe=skip:error` — network error during probe, request allowed to proceed
- `repository_permission_denied` in the audit log — the user lacks required read/write permission for a service-PAT call
+55
@@ -0,0 +1,55 @@
# Write Mode
## Threat Model
Write mode introduces mutation risk (issue/PR changes, metadata updates). Risks include unauthorized action, accidental mass updates, and audit evasion.
## Default Posture
- `WRITE_MODE=false` by default.
- When enabled, writes require repository whitelist membership by default.
- Optional opt-in: `WRITE_ALLOW_ALL_TOKEN_REPOS=true` allows writes to any repo the token can access.
- Policy engine remains authoritative and may deny specific write tools.
## Supported Write Tools
- `create_issue` (optional `milestone` id or title)
- `update_issue` (optional `milestone`; `0` clears it)
- `create_issue_comment`
- `create_pr_comment`
- `edit_issue_comment`
- `add_labels`
- `remove_labels`
- `assign_issue`
- `create_label`
- `update_label`
- `create_pull_request`
- `create_release`
- `edit_release`
- `create_branch`
- `create_milestone`
Not supported (explicitly forbidden): merge actions, branch/label/release deletion,
force push, repo/admin management, and repository content writes (file create/edit,
commits). Gitea Projects (Kanban boards) are unsupported because the Gitea REST API
exposes no project endpoints.
## Enablement Steps
1. Set `WRITE_MODE=true`.
2. Choose one:
- `WRITE_REPOSITORY_WHITELIST=owner/repo,...` (recommended)
- `WRITE_ALLOW_ALL_TOKEN_REPOS=true` (broader scope)
3. Review policy file for write-tool scope.
4. Verify audit logging and alerting before rollout.
## Safe Operations
- Start with one repository in whitelist.
- Use narrowly scoped bot credentials.
- Require peer review for whitelist/policy changes.
- Disable write mode during incident response if abuse is suspected.
## Risk Tradeoffs
Write mode improves automation and triage speed but increases blast radius. Use least privilege, tight policy, and strong monitoring.
+1
@@ -0,0 +1 @@
_Generated from the docs/ directory. Edit the docs, not the wiki, then re-run the wiki sync._
+27
@@ -0,0 +1,27 @@
### AegisGitea-MCP
**Start**
- [Home](Home)
- [Getting Started](Getting-Started)
- [Configuration](Configuration)
**Operating**
- [Write Mode](Write-Mode)
- [Automation](Automation)
- [Deployment](Deployment)
**Internals**
- [Architecture](Architecture)
- [Policy](Policy)
- [Audit](Audit)
- [Observability](Observability)
**Security**
- [Security](Security)
- [Hardening](Hardening)
- [Governance](Governance)
**Reference**
- [API Reference](API-Reference)
- [Troubleshooting](Troubleshooting)
- [Roadmap](Roadmap)