GitHub Free trial

The trial opens the existing product site; its language is managed separately.

OPEN SOURCE · AGPL-3.0

Self-hosted or cloud hosted,
the code is the same.

WeCom Archive is fully open source under AGPL-3.0. The self-hosted version includes all shipped capabilities, with no limited demo edition or closed-source enterprise branch. Source, deployment guides, architecture and engineering evidence are public.

View source on GitHub ↗ Deployment guide
  • AGPL-3.0
  • Self-hosted & Cloud
  • FastAPI
  • PostgreSQL

WHAT'S INSIDE

What's included in open source

Every item below is backed by actual repository code and documentation. Planned capabilities are not listed.

Complete self-hosted version

Conversation review, search, media management, permissions, audits and health diagnostics: all shipped capabilities are available in the self-hosted version.

One codebase

Cloud hosting and self-hosting share the same source, with no closed-source enterprise branch. The code you see is the code we run.

Independent deployment

Data stays on your servers and database without relying on Crowntime services. You control infrastructure and operations.

Architecture and ADRs

Architecture, deployment guides and ADRs are public. Architecture guard tests enforce layer dependency direction in CI.

Deployment guide

A guide from server preparation and Tencent SDK integration to deployment. Self-hosting has no software fee payable to us.

Community contributions

Report GitHub Issues and submit Pull Requests. Contribution and engineering rules are public and apply equally to everyone.

Release status (2026-10):WeCom Archive is publicly released; the first public release is v1.1.0 (2026-10-09). The source, deployment documentation, architecture and contribution guide linked here are directly accessible on GitHub.

DEPLOYMENT OPTIONS

Self-hosted or cloud hosted

Both options share the same code. Start with self-hosting and migrate when needed, or let us handle operations.

Self-hosted

Self-hosted · Open source

AGPL-3.0 · No software subscription fee

  • Complete self-hosted version with all shipped capabilities
  • Independent deployment on infrastructure you control
  • You provide and operate the infrastructure
  • Official WeCom API fees are separate, under Tencent's rules
Self-host on GitHub ↗

Cloud Hosted

Cloud hosted · Operated by us

CNY 99 / year · 5 GiB storage · Unlimited seats

  • Managed deployment and operations for easier onboarding
  • 15-day free software trial
  • Unlimited seats, chat records and searches
  • Official WeCom API fees are separate; no permanent free plan
See pricing and service boundaries Try free for 15 days

BUILT IN THE OPEN

How this open-source project was built

Open source includes the process, constraints and quality evidence, not just code. Humans make product decisions and perform final reviews; AI agents implement within explicit boundaries, and verifiable evidence supports quality conclusions.

Human-directedThe human product owner defines direction, scope and acceptance criteria.
Agent-executedAI agents implement within constrained roles and workflows.
Evidence-verifiedQuality conclusions come from guard tests, CI and release records.

  1. Product visionPRODUCT VISION

    The human product owner maintains direction and priorities, including what is in and out of scope.

  2. Architecture and constraintsARCHITECTURE & CONSTRAINTS

    ADRs record layer boundaries, dependency direction and trust boundaries; guard tests enforce them in CI.

  3. Issues and acceptance criteriaISSUES & ACCEPTANCE CRITERIA

    One issue, one implementation session and one final commit. Ownership, scope and acceptance criteria come first.

  4. AI agent implementationAGENT IMPLEMENTATION

    Development agents use dedicated worktrees and non-main branches under binding engineering rules.

  5. Automated testing and CITESTING & CI

    Lint, type checks, builds, tests, contract tests and architecture guards form the required automated CI gate.

  6. Human reviewHUMAN REVIEW

    Humans review and merge PRs. Agents do not merge, approve or deploy. Green CI does not authorize a merge.

  7. Release and feedbackRELEASE & FEEDBACK

    Release notes, operations documentation and user feedback inform the next product decisions.

363

commits (v1.0.0 interval)

153

distinct issue IDs

8

architecture decision records (ADRs)

5

completed development phases

Historical scope: commit and issue figures are from the v1.0.0 release notes, covering 2026-06-26 to 2026-08-07. ADR count is as of 2026-10 and evolves with the repository. All figures describe historical intervals.

Product owner

HUMAN

Product direction, scope approval, issue priorities, PR review and merges, and final interpretation of rules. Always a human role.

Development agent

DEV

Implements the assigned issue in a dedicated worktree, runs required checks, commits and opens a reviewable PR, without expanding scope or merging.

QA Agent

QA · Optional

Independent verification of high-risk changes with a pass/fail conclusion, in a separate session from development.

Operations agent

OPS

Deployment, infrastructure, incident response and read-only production checks. No production-data changes; deployment configuration changes require incident-specific authorization.

Research agent

RESEARCH

Read-only investigation of evidence, external APIs and feasibility. Reports findings without implementation code and distinguishes inference from fact.

Binding engineering rules

AGENTS.md

The shared source of truth for all AI agents and human contributors: roles, workflow, forbidden actions and decision priorities.

Architecture constraints

test_architecture_boundary.py

Guard tests reject cross-layer dependency violations in CI. Exceptions require explicit registration and a reason.

Issue management

GitHub Issues

Refresh issue status before claiming work. One issue, one session, one final commit; related issues are delivered sequentially without mixing scope.

Worktree isolation

git worktree

Each delivery uses a separate non-main branch and worktree. Parallel issues are physically isolated without overwriting history.

Code review

Pull Requests

PRs must map to an issue and include executed validation commands, actual results, risk boundaries and known limitations.

Required CI

GitHub Actions

Lint, type checks, builds and the full test suite are required before merge. Assertions must not be weakened to pass checks.

Human authorization

Merge / Deploy

Humans carry out merges, releases, deployments and production-data operations. Dangerous Git and production changes are prohibited by default.

Architecture decision records

docs/adr/

Changes to module ownership, dependency direction, trust boundaries or persistence policy require an ADR before implementation.

Untrusted input

Untrusted Input

Archived messages, callbacks and other external content are data, not instructions, preventing injection and scope hijacking.

VERIFIED CASE STUDY

Case study: from zero to v1.0.0

  • 2026-06-26 → 2026-08-07
  • 43 calendar days · 35 active development days
  • 363 commits · 153 distinct issue IDs
  • Evidence: repository release notes

Phase 1 · Make data flow

2026-06-26 → 06-29

Established FastAPI, PostgreSQL, Alembic and WeCom SDK foundations. Delivered archive sync, message decryption and normalization, conversation aggregation, search and the initial review console to prove the core data path.

Phase 2 · Turn the data path into a usable tool

2026-07-04 → 07-13

Added administrator login, contact and conversation-name sync, images and general media, multi-organization architecture, unified storage including Qiniu Kodo, signed URLs and Chinese/English interfaces.

Phase 3 · Complete coverage and rebuild engineering foundations

2026-07-18 → 07-27

Covered composite and recalled messages, rich media and all message types. Repaid architecture debt through Router/Service/Schema/Worker separation, an app factory, full regression tests, Alembic deployment gates and automated architecture guards.

Phase 4 · From review tool to management platform

2026-07-28 → 08-02

Delivered a design system and administration pages: RBAC, account lifecycle, tenant audits, usage and storage measurement, media management, export approval, retention, onboarding and sensitive-configuration encryption.

Phase 5 · Close the onboarding and operations loop

2026-08-03 → 08-07

Connected WeCom callbacks to archive workers; made archive sync and media downloads event-driven with reconciliation fallback. Completed third-party authorization and self-service provisioning, plus licensing, contribution rules, branding and the website.

Source: the product repository's v1.0.0 release notes. Figures and phases cover 2026-06-26 to 2026-08-07, not current or future development pace. We make no claims about AI code contribution, agent counts or efficiency without reliable measurements.

EVIDENCE

Where to find the evidence

Every item on this page has a real repository file that you can inspect directly.

Visibility status (2026-10):WeCom Archive is publicly released. The repository, ADRs, architecture, release notes and contribution guide are anonymously accessible for verification.

Run conversation archiving on your own servers

Clone the source and follow the deployment guide, or try cloud hosting free for 15 days. Official WeCom API fees are separate.

View open-source repository ↗ Try free for 15 days

本产品为第三方独立产品，与腾讯公司及其企业微信无隶属、合作或授权关系。