MVP pilot · bugAgent-maintained community integration

bugAgent for Hermes Agent

Run curated bugAgent test plans with the agent runtime you choose, then return every result to one governed QA record.

Connect through MCP Use the REST API Get the public starter kit

This community integration is maintained by bugAgent. It is not maintained, endorsed, or supported by Nous Research.

One controlled execution loop

The agent executes. bugAgent governs.

Hermes provides the runtime, model, and browser tooling. bugAgent keeps the approved plan, tenant boundary, immutable snapshots, canonical results, and human review.

01 Curate

A human approves a project-bound suite in bugAgent.

02 Execute

Hermes retrieves a bounded snapshot and exercises the target.

03 Record

Retry-safe batches update one canonical run and summary.

04 Review

A person reviews failures and owns defect and release decisions.

Pilot setup

Connect a non-production suite

  1. 1
    Create a least-privilege key

    Grant only test_runs:read and test_runs:write. The key is bound to its bugAgent workspace.

  2. 2
    Add bugAgent MCP to Hermes

    Allow only the suite listing and agent execution tools needed by the skill.

  3. 3
    Run a synthetic pilot

    Use a staging target, a test account, and a suite created for integration testing before scheduling runs.

MCP setup and tools REST contract Public Hermes skill
~/.hermes/config.yaml
mcp_servers:
  bugagent:
    url: "https://mcp.bugagent.com/mcp"
    headers:
      Authorization: "Bearer ${BUGAGENT_API_KEY}"
    tools:
      include:
        - list_test_suites
        - start_test_plan
        - get_test_run_plan
        - report_test_results
        - abort_test_run

First run

Run your first governed suite

Install the public community skill, connect the scoped key, and name one approved suite and project in the prompt. Hermes executes the immutable plan; bugAgent remains the system of record.

  1. Create the scoped keyIn Developer settings, grant only test_runs:read and test_runs:write.
  2. Install and verifyReview the skill, install it from GitHub, and test the MCP connection.
  3. Execute a safe targetUse a staging environment and a synthetic login with externally restricted browser and network access.
  4. Review in bugAgentConfirm every case appears once and the pass, fail, blocked, and skipped totals match.
Install and connect
hermes skills install TestLauncher/testlauncher/skills/bugagent-qa
hermes skills list

# Add the real key only to your local Hermes environment.
printf '%s\n' 'BUGAGENT_API_KEY=ba_live_REPLACE_LOCALLY' >> ~/.hermes/.env
chmod 600 ~/.hermes/config.yaml ~/.hermes/.env
hermes mcp test bugagent
Hermes prompt
/bugagent-qa Execute the "Checkout smoke" suite in the API Testing project.
Use the staging target already configured for this Hermes profile. Do not make
purchases, invite users, or perform destructive actions. Report every result to
bugAgent and abort safely if the target cannot be reached.

Replace the example suite and project names with your own. Never put target credentials or a live API key in the prompt.

Overlap evaluated honestly

Use each product for what it does best

Both can participate in test authoring and execution. The integration avoids duplicating a general agent runtime inside bugAgent or moving the QA system of record into an execution worker.

Area Hermes Agent bugAgent
RuntimeModel choice, skills, memory, scheduling, browser and computer toolsPurpose-built QA workflows and managed test services
ExecutionFlexible customer-controlled workerCurated scope, immutable plan, canonical run and evidence record
GovernanceDepends on the customer's deployment and toolsWorkspace and project isolation, API scopes, audit trail, retesting lifecycle
DecisionChoose Hermes when runtime freedom and self-hosting matter mostChoose bugAgent when QA traceability, review, and one system of record matter

Plan and runtime boundary

Know the limits before you run

  • Free includes 10 stored test cases, 1 suite, 3 folders, and 10 total test runs per UTC calendar month.
  • Each Free test case has a 128 KB structured-content maximum. Free can use URL references; test-case file attachments require Team or Enterprise.
  • Up to 3 Free runs per month may use Hermes or another external agent. Free allows 1 active external run and at most 10 cases in each external plan.
  • Free allows 2 active workspace API keys, so Hermes can use a dedicated test_runs:read/test_runs:write key without sharing a general integration credential.
  • Free execution traffic is limited to 30 requests per key and 60 per workspace per minute. Team and Enterprise allow 120 per key and 600 per workspace per minute.
  • Retrying with the same external_run_id resumes the existing run and does not consume another run. Deleting data does not reset monthly run usage.
  • Completed and aborted external runs are final and cannot be reopened to recycle quota.
  • Team and Enterprise test case storage and runs are unlimited, subject to general platform protections. AI test-case generation, AI tag suggestions, and Figma import also require Team or Enterprise.
  • It executes approved suites; it does not silently expand test coverage.
  • Result evidence is text in the MVP; agent attachment upload comes later.
  • Target browser, model, and network costs remain customer-side, as do browser reliability and target authentication.
  • Test case text is untrusted input; customers must restrict Hermes tools, target access, and network egress.
  • Target credentials stay in the execution runtime and never enter the bugAgent plan contract.
  • A human still decides whether to file a bug or approve a release.

Private pilot

Pilot one staging suite

Measure setup time, completion rate, result quality, and repeat use before expanding the integration.

Create a scoped key