Your team and coding agents, working as one.

Markbase gives development teams the shared knowledge, coordination, and continuity they need to build, operate, and evolve software with coding agents.

Start free

Copies a setup prompt for your coding agent
Coverdash

We chose Markbase’s on-prem deployment to give our team a shared knowledge base while keeping control of our data and infrastructure. We’re excited to bring it into our workflows at Coverdash.

Perry RaskinHead of Web & Integrations, Coverdash

Knowledge

[ 01 / 03 ]

Long-term shared memory layer for your team and its agents.

Shared knowledge

Agents start from what your team knows and add what they learn.

Each task begins with the team's decisions, past incidents and notes, and what the agent finds goes back into Markbase, so the next task starts further ahead.
decisions kept
86
decisions revised
14
findings recorded
128
changes shipped
57

Shared knowledge

Agent task

Written back

Shared skills

When someone improves a skill, the whole team gets it.

Someone improves the team's code-review skill so agents flag changes that arrive without tests. Everyone on the team gets the new version, whichever coding agent they use.
code-reviewa3f9c1
Cursor

Ana

a3f9c1
Claude Code

Leo

a3f9c1
Gemini

Priya

a3f9c1
Codex

Sam

a3f9c1
GitHub Copilot

Jo

a3f9c1

Editing together

Several agents can edit the same knowledge at once.

Nothing is locked and nobody has to sort out a merge. Each agent changes only the part it means to change, so edits made at the same time all land. If another agent changed that part first, Markbase tells it, and it reads the new version before making its change.

Inboxes

Forward an email to your agent and it can read it.

When an email holds what your agent needs, like a customer's bug report or a failed payment, ask your agent for an address. It makes one that only takes mail from you and expires on its own. Forward the email, and the agent reads it with its attachments, without access to your mailbox.

Inboxes provided by SendOps

k7q2m9@sndps.comonly from Priya · expires in 30 min

Priya

Claude Code

From A customer

Fwd: Checkout fails on Safari

screenshot.png
Typed records
Keep decisions, incidents or anything else your team tracks as structured records, so agents can find the right ones quickly.
Version history
Markbase keeps earlier versions, so you can see what changed, go back to an older one, or lock something so agents can't edit it.

Continuity

[ 02 / 03 ]

Pick up long-running work where you left off.

Follow-ups

Park work until it's due, then pick it up with its context.

An agent prepares the rollout plan at beta launch and parks it. When the trial ends, a runner with read-only production access picks it up to check the results, and at GA release a person reviews the evidence and makes the call.

Beta launch

Agent prepares

Trial ends

Runner resumes

GA release

Human reviews

Handoffs

Pass a task to the agent and machine that have the access it needs.

Markbase queues a job to reproduce a failed checkout. Codex pulls it on a CI runner that works unattended and has no route to production, then writes up what it found. The next step needs production logs, so Markbase queues it with the write-up. Claude Code pulls it from a runner inside the team's Tailscale network, which needs no inbound access, and asks Ana before it runs each query.

Markbase

waiting

ci-runner

Codex

  • open internet
  • auto mode
waiting

ops-runner

Claude Code

  • tailnet
  • asks first
waiting
Due dates and reminders
Put a date on work you've paused, with a note on how to pick it up. Markbase reminds you when it's due and hands you the note to give your agent.

Automation

[ 03 / 03 ]

Start an agent when something happens in your other tools.

Events

When an alert comes in, an agent investigates it with the project's history.

An event from Sentry, PagerDuty, Linear or any service that sends a webhook reaches Markbase, which starts Claude Code with the project's past incidents and decisions. It reproduces the failure, writes a fix and opens a pull request.

Sentry

Markbase

Claude Code

Pull request

Git-aware knowledge

Knowledge written on a branch stays off main until the pull request merges.

An agent writes the design and decisions for queue retries on a branch, linked to its pull request. Until it merges, an agent searching from main doesn't see them. When GitHub, GitLab or Bitbucket reports the merge, Markbase marks just that knowledge shipped, in seconds. If the pull request is closed instead, it is marked abandoned.

main

feat/queue-retries

Design

Decision

Runbook

Agent on main

Code to knowledge

When the code changes, an agent updates what Markbase says about it.

Most tools only turn a written spec into code. Markbase also works the other way. When a merged pull request changes how checkout retries failed payments, Markbase starts an agent to find what still describes the old behaviour, such as the architecture, the runbook or the changelog, and bring it up to date.

checkout.ts

merged

Architecture
Your own prompts
You decide what the agent is asked to do for each kind of event, and you can check the prompt before it runs.
Runs on your machines
Agents run on the machines you choose, with only the access you give them, and you don't have to open your network to Markbase.

[ Works with ]

Markbase works with the tools you already use.

Your coding agents share what they know through Markbase, and your other tools can start an agent when something happens. Anything that can send a webhook works too.

  • Claude Code
  • GitHub
  • Sentry
  • Codex
  • Linear
  • Cursor
  • PagerDuty
  • Gemini
  • Slack
  • GitLab
  • Grok
  • incident.io
  • Stripe
  • DeepSeek
  • Bitbucket
  • Webhooks
  • Kimi
  • Shopify
  • Axiom
  • Svix

[ Use cases ]

Use cases