Baltor Showcase
baltor.ai

Watch a big job run, step by step.

Pick a job, read its checked steps and press Run, all in your browser before you connect a model key.

Choose a job

Execution steps

{{progress}}
  1. Compare MCP patterns and services{{st1.label}}

    Check: each option names its sourceFeeds

  2. Pull a server skill and contract schemas{{st2.label}}

    Check: checksum matchesPro

  3. Generate tools from your API spec, with tests{{st3.label}}

    Check: every tool has a passing testPro

  4. Add MCP server settings{{st4.label}}

    Check: settings load and list each toolPro

  5. Connect it to Claude Code, Codex, OpenCode and Pi{{st5.label}}

    Check: each agent lists the toolsPro
  1. Read the release notes and breaking changes{{st1.label}}

    Check: each change links its release noteFeeds

  2. Pull the migration skill, codemods and test hook{{st2.label}}

    Check: checksums matchPro

  3. Apply one change set per step, tests after each{{st3.label}}

    Check: tests pass after each change setPro

  4. Queue the long tail overnight{{st4.label}}

    Check: limits and checkpoints setMax

  5. Review the morning report{{st5.label}}

    Check: each result passed before the next stepMax
  1. Read security sources for your stack{{st1.label}}

    Check: each source carries its dateFeeds

  2. Pull the review subagent, scanning hook and audit command{{st2.label}}

    Check: checksums matchPro

  3. Review findings step by step{{st3.label}}

    Check: each finding names a file and linePro

  4. Fix the clear ones, flag the rest for you{{st4.label}}

    Check: tests pass after each fixPro

  5. Read the morning report{{st5.label}}

    Check: every flag reaches the reportMax
Run output Sample output

Turn your internal API into an MCP server

You hand offopenapi.yaml for your orders service You get backan MCP server with typed tools, a test for each tool, and settings for Claude Code, Codex, OpenCode and Pi Plan5 steps, each checked before the next begins

Press Run to start step 1.

Step 1 of 5: Compare MCP patterns and services

Agent FeedsChoose MCP tools and servicesDiscover services an agent can connect to and inspect permissions before enabling them. Comparedone tool per endpoint
tools grouped by resource
one gateway server for every service Chosentools grouped by resource, reads and writes kept apart
Check  each option names its sourcePassed

Step 2 of 5: Pull a server skill and contract schemas

intelligence_search query "MCP server from an OpenAPI spec"
1openapi-to-mcp-serverskill, MIT, 6.4 KB, sha256 3f9c1a07…Chosen
2mcp-tool-schema-checkscode module, Apache-2.0, 2.2 KB, sha256 b81e44d2…Chosen
provisioning_read identity openapi-to-mcp-server expected_digest 3f9c1a07…
Check  checksum matchesPassed

Step 3 of 5: Generate tools from your API spec, with tests

Readopenapi.yaml, 14 operations Wroteserver/tools/orders_list.ts
server/tools/orders_get.ts
server/tools/orders_create.ts
and 11 more tools Tests14 written, 14 passed
Check  every tool has a passing testPassed

Step 4 of 5: Add MCP server settings

.mcp.json
{
"mcpServers": {
"orders": {
"command": "node",
"args": ["server/index.js"]
}
}
}
Check  settings load and list 14 toolsPassed

Step 5 of 5: Connect it to Claude Code, Codex, OpenCode and Pi

claude-codeconnected, 14 tools codexconnected, 14 tools opencodeconnected, 14 tools piconnected, 14 tools

Result: an MCP server for your orders service, ready in four agents.

Check  each agent lists the same 14 toolsPassed

Upgrade an app to the next major framework version

You hand offyour app on framework version 4 You get backthe app on version 5, one tested change set at a time, and a morning report on the long tail Plan5 steps, each checked before the next begins

Press Run to start step 1.

Step 1 of 5: Read the release notes and breaking changes

Agent FeedsPlan platform and tool upgradesFind relevant releases, breaking changes and deprecations for a chosen technology stack. Found7 breaking changes that touch your code First threerouting imports moved
config file renamed
data loader signature changed
Check  each change links its release notePassed

Step 2 of 5: Pull the migration skill, codemods and test hook

intelligence_search query "framework 5 migration codemods"
1framework-5-migrationskill, MITChosen
2codemod-routing-importscode module, MITChosen
3test-after-changehook, Apache-2.0Chosen
provisioning_read 3 items, each with its expected_digest
Check  checksums matchPassed

Step 3 of 5: Apply one change set per step, tests after each

Change set 1routing imports, 142 files, tests passed Change set 2config file, 1 file, tests passed Change set 3data loaders, 38 files, tests passed Left for laterchange sets 4 to 7
Check  tests pass after each change setPassed

Step 4 of 5: Queue the long tail overnight

Queuedchange sets 4 to 7, one per step Call limit400 model calls Time limit6 hours Checkpointsafter every change set Each stepstarts in a fresh agent session, clean and focused
Check  limits and checkpoints setPassed

Step 5 of 5: Review the morning report

Merged2 change sets Ready for review1 change set Waiting for you1 question about a custom plugin Calls used186 of 400

Result: the app on version 5, with one question waiting for you.

Check  each result passed before the next stepPassed

Run a security pass before release

You hand offthe release branch You get backfindings by file and line, the clear ones fixed and the rest flagged for you Permissionsthe ones you choose, per step Plan5 steps, each checked before the next begins

Press Run to start step 1.

Step 1 of 5: Read security sources for your stack

Agent Feedsadvisories, release notes and dependency notices for your stack Read12 sources, each with its date Watch list2 advisories that match your lockfile
Check  each source carries its datePassed

Step 2 of 5: Pull the review subagent, scanning hook and audit command

intelligence_search query "security review before release"
1security-reviewsubagent, Apache-2.0, reads files, writes a reportChosen
2scan-staged-secretshook, MIT, reads staged filesChosen
3dependency-auditcommand, MIT, reads the lockfileChosen
Check  checksums matchPassed

Step 3 of 5: Review findings step by step

Finding 1tests/fixtures/env.sample:4a live-looking key in a test fixture Finding 2package.json2 dependencies with published advisories Finding 3api/upload.ts:88file type taken from the client
Check  each finding names a file and linePassed

Step 4 of 5: Fix the clear ones, flag the rest for you

FixedFinding 1key moved to an environment variable name FixedFinding 2dependencies raised to patched versions FlaggedFinding 3your call on which file types to accept
Check  tests pass after each fixPassed

Step 5 of 5: Read the morning report

Fixed, merged2 findings Flagged for you1 finding, with its file, line and a suggested fix Permissionsread files, write in the step folder, run tests

Result: a release branch with two fixes in and one decision waiting for you.

Check  every flag reaches the reportPassed
Every output here is a labelled example, with illustrative item names.

Recorded runs, too.

The sandbox shows the shape of a run. These case studies show real ones, every result included.

Each step uses one part of Baltor.

Research, build and keep working. Bring your own model access; your model keys stay with you.

Research

Agent Feeds

Source collections your agent reads before it decides.

$4.99a month, free through December 31, 2026 (Eastern)

Build

Pro, with Agent Feeds + Harness Files

Skills, tools, code, hooks and settings, at exact versions with checksums.

$29a month

Keep working

Max, with Agent Feeds + Harness Files + Overnight / AFK Work

Queues, call limits, checkpoints and a morning report.

$49a month, Public preview