Techlog

Connect context across🤖 agents and 💬 sessions.

Your AI agent records the problem, the decision, the change, the tests, and the incident in Techlog as it works. Reviews get shorter, incidents get mitigated faster, and every postmortem ends in action.

Proposed standard · draft 0.1 Fictional, simulated data Local server runs No hosted service yet

Legacy Changelog

  • 1 · Problem💬 Slack“Why a 20% cap?”
  • 2 · Decision🎫 JiraSPIN-214 · cap approved
  • 3 · Change🔀 GitHubPR #482 · spin-to-win
  • 4 · Release🚀 GitHubRelease 3.8.0 · canary
  • 5 · Incident🚨 PagerDutyP1 · 5xx at 6.2%
  • 6 · Rollback💬 Slack“Rolling back, I think?”
  • 7 · Postmortem📄 ConfluencePostmortem 2026-03-12
  • 8 · Corrective action🎫 JiraOPS-731 · 6 actions
  • 🤖 AI agentClaudeOpens PR #482
  • 🤖 AI agentCodexWrites the postmortem

Humans open one story and click from problem to decision, change, release, incident, and postmortem.

AI agents ask over MCP and get the same linked context, not five tabs.

Try Techlog Read the docs

How it works

An agent reports progress. Techlog connects it.

API first: every step is a call, and the website is just another client.

  1. 1

    Agents confirm

    Agents help convert hypotheses into confirmed facts, with evidence.

  2. 2

    Changelog records

    Each step is one record, sent as an MCP request: techlog_record_event.

  3. 3

    Build connections

    Links the record to the problem, decision, tests, and release.

  4. 4

    Store, serve

    One event store, read by API, MCP, and the UI.

Open standard · draft 0.1

The event is plain JSON.

Any agent or tool can write it, and any tool can read it. A small required core, typed links marked confirmed or hypothesis, and a record of who produced it. Below, a techlog_record_event call with the ids of the example story, abridged to the fields that matter.

Event model Read the standard →
Request, abridged
{
  "jsonrpc": "2.0",
  "id": 7,
  "method": "tools/call",
  "params": {
    "name": "techlog_record_event",
    "arguments": {
      "event": {
        "idempotency_key": "seed-spring-spin-004",
        "type": "proposal.recorded",
        "actor": { "kind": "ai_agent", "id": "agent:code-assistant" },
        "subject": { "system": "promo-api", "component": "spin-service" },
        "status": "proposed",
        "summary": "Spin-to-win: one spin per new account, server-side draw",
        "links": [
          { "rel": "motivated_by", "target": "evt_01J9K2A0P002", "basis": "confirmed" }
        ]
      }
    }
  }
}

Available connectors

Connect the tools you already use.

Proposed · none available yet

GitHub

Source and CI

Pull requests, commits, and workflow runs become linked events.

Slack

Collaboration

Decisions and approvals from chat stop disappearing.

Atlassian

Jira, Confluence, Service Management

Issues, incidents, and write-ups join the same story.

Your tool

Anything that writes an event

The event model is open, so no permission is needed.

Ask for a connector Build your own →

GitHub, Slack, Atlassian, Jira, Confluence, and PagerDuty are trademarks of their owners, named only to show what a connector would link to or what teams use today. No partnership or availability is implied.

Run it on your machine today.

A basic Techlog server runs locally with the example story loaded. Or read how the event model works first.

Try Techlog Read the docs