When making a POST request to /hooks/agent, the chat history for the same session will be cleared. | Friends of the Crustacean 🦞🤝 | Page 1

wispy perch Apr 29, 2026, 3:16 AM

#

curl -X POST https://your-gateway/hooks/agent
-H "Authorization: Bearer your-secret-token"
-H "Content-Type: application/json"
-d '{
"agentId": "my-agentId",
"message": "",
"sessionKey": "optional-custom-session"
}'

wispy flowerBOT Apr 29, 2026, 3:16 AM

#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

limpid thicketBOT Apr 29, 2026, 3:18 AM

#

Please share openclaw --version first — behavior here has changed over time.

On current docs/source: /hooks/agent is an isolated hook run, not a normal continuing chat session. A stable sessionKey scopes/routs the hook, but the run is currently forced fresh internally, so previous transcript/history won’t be reused like a regular chat thread.

Also: your example has "message": "", which should be rejected with message required; it needs a non-empty message.

If you want stable context:

for main-session wakeups, use POST /hooks/wake
for /hooks/agent, set hooks.defaultSessionKey / hooks.allowRequestSessionKey only for routing/scoping, not guaranteed transcript continuation
if you expected /hooks/agent to preserve history across same sessionKey, that’s worth filing as a GitHub issue with your version + repro: https://github.com/openclaw/openclaw/issues

Docs: https://docs.openclaw.ai/gateway/configuration-reference#hooks 🦐

wispy perch Apr 29, 2026, 3:25 AM

#

limpid thicket Please share `openclaw --version` first — behavior here has changed over time. ...

My workflow is as follows: I start my automated tests via session A using OpenClaw. After the tests are completed, I request a webhook to notify OpenClaw that the tests are finished and begin analyzing the test report. So, how should I make this request?

limpid thicketBOT Apr 29, 2026, 3:26 AM

#

For your workflow, I’d use /hooks/wake, not /hooks/agent, if you want OpenClaw to continue from the main/session-A conversation context.

Example:

curl -X POST https://your-gateway/hooks/wake \
  -H "Authorization: Bearer your-hooks-token" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: test-run-123-finished" \
  -d '{
    "mode": "now",
    "text": "The automated test run I started is finished. Please analyze the test report at /path/to/report.xml and summarize failures, likely causes, and next actions."
  }'

Use /hooks/agent only if you’re okay with a fresh isolated analysis run. In that case, make the message fully self-contained:

#

curl -X POST https://your-gateway/hooks/agent \
  -H "Authorization: Bearer your-hooks-token" \
  -H "Content-Type: application/json" \
  -d '{
    "agentId": "my-agentId",
    "name": "Test report analysis",
    "message": "Analyze this completed test report: /path/to/report.xml. Summarize failures and suggested fixes.",
    "deliver": true,
    "wakeMode": "now"
  }'

So the short version:

Continue the existing OpenClaw conversation / session A: use /hooks/wake
Run a separate webhook-triggered agent job: use /hooks/agent

#

Don’t rely on /hooks/agent sessionKey for chat-history continuation; treat it as routing/scoping.

Docs: https://docs.openclaw.ai/automation/cron-jobs#webhooks 🦐

wispy perch Apr 29, 2026, 3:29 AM

#

limpid thicket For your workflow, I’d use **`/hooks/wake`**, not `/hooks/agent`, *if you want O...

If using /hooks/wake, how do I send it to the corresponding session?

limpid thicketBOT Apr 29, 2026, 3:32 AM

#

Short answer: you can’t target a specific session with /hooks/wake right now. It only enqueues a system event for the main session.

If you need “send this completion event back into session A”, use a session-routed Gateway call instead, e.g. OpenAI-compatible HTTP:

curl -sS https://your-gateway/v1/chat/completions \
  -H "Authorization: Bearer $OPENCLAW_GATEWAY_TOKEN" \
  -H "Content-Type: application/json" \
  -H "x-openclaw-session-key: agent:main:discord:..." \
  -d '{
    "model": "openclaw/default",
    "messages": [{
      "role": "user",
      "content": "The test run is finished. Analyze the report at /path/to/report.xml and summarize failures."
    }]
  }'

#

Important differences:

/hooks/wake = main session only.
/hooks/agent = isolated run, not session-A history.
/v1/chat/completions + x-openclaw-session-key = explicit session routing.
This uses the Gateway token, not the hooks token, so keep it on loopback/tailnet/private ingress only.

Docs: https://docs.openclaw.ai/gateway/openai-http-api
Session keys: openclaw sessions --json 🦐

wispy perch Apr 29, 2026, 5:10 AM

#

limpid thicket Short answer: **you can’t target a specific session with `/hooks/wake` right now...

My requirement is to notify OpenClaw to start analyzing the test report after the automated test is completed, and then notify a specific agent to modify the test script after the analysis is finished.

limpid thicketBOT Apr 29, 2026, 5:12 AM

#

Yep — for that workflow I’d not use /hooks/wake. Use a small external orchestrator:

Test runner finishes.
Call OpenClaw to analyze the report.
When that HTTP call returns, call the specific “modifier” agent with the analysis.

Best fit is the Gateway OpenAI-compatible endpoint because it’s request/response:

# 1) analyze
ANALYSIS=$(
  curl -sS "$GATEWAY/v1/chat/completions" \
    -H "Authorization: Bearer $OPENCLAW_GATEWAY_TOKEN" \
    -H "Content-Type: application/json" \
    -H "x-openclaw-session-key: agent:test-analyzer:ci-$BUILD_ID" \
    -d '{

#

      "model": "openclaw/test-analyzer",
      "messages": [{
        "role": "user",
        "content": "Tests completed. Analyze the report at /path/to/report.xml. Return root cause and recommended script changes."
      }]
    }' | jq -r '.choices[0].message.content'
)

# 2) ask modifier agent to change the script
jq -n --arg analysis "$ANALYSIS" '{
  model: "openclaw/test-modifier",
  messages: [{
    role: "user",
    content: ("Use this analysis to modify the test script:\n\n" + $analysis)
  }]

#

}' | curl -sS "$GATEWAY/v1/chat/completions" \
  -H "Authorization: Bearer $OPENCLAW_GATEWAY_TOKEN" \
  -H "Content-Type: application/json" \
  -H "x-openclaw-session-key: agent:test-modifier:ci-'$BUILD_ID'" \
  -d @-

Important bits:

Enable gateway.http.endpoints.chatCompletions.enabled.
Use model: "openclaw/<agentId>" to target a specific agent.
Use x-openclaw-session-key if you want stable session history.
/hooks/agent is more fire-and-forget: it returns a runId, not the completed analysis, so it’s weaker for “after analysis finishes, then do X”.

Docs:
https://docs.openclaw.ai/gateway/openai-http-api

#

https://docs.openclaw.ai/automation/cron-jobs

If you’re seeing history cleared specifically with /hooks/agent, also share openclaw --version — session-key behavior is version-sensitive.

wispy perch Apr 29, 2026, 7:57 AM

#

limpid thicket Short answer: **you can’t target a specific session with `/hooks/wake` right now...

A Postman request to http://127.0.0.1:18789/v1/chat/completions returned a 404 error.

limpid thicketBOT Apr 29, 2026, 7:57 AM

#

A 404 there usually means the OpenAI-compatible HTTP endpoint is not enabled yet, or you’re hitting a Gateway version that doesn’t have it.

First, please check:

openclaw --version

Then enable this in your OpenClaw config:

{
  gateway: {
    http: {
      endpoints: {
        chatCompletions: { enabled: true },

#

      },
    },
  },
}

Restart the gateway:

openclaw gateway restart

Smoke test:

#

curl -sS http://127.0.0.1:18789/v1/models \
  -H "Authorization: Bearer $OPENCLAW_GATEWAY_TOKEN"

Then retry POST /v1/chat/completions.

Also make sure Postman is using POST, with:

Authorization: Bearer <gateway token>
Content-Type: application/json

Docs: https://docs.openclaw.ai/gateway/openai-http-api

wispy perch Apr 30, 2026, 1:12 AM

#

limpid thicket A 404 there usually means the OpenAI-compatible HTTP endpoint is **not enabled**...

Sometimes OpenClaw receives OpenAI's chat completion requests but doesn't respond. Why is this? (openclaw 2026.4.7)

limpid thicketBOT Apr 30, 2026, 1:13 AM

#

On 2026.4.7, the most likely reason is: non-streaming /v1/chat/completions only responds after the whole OpenClaw agent turn finishes.

So if the selected agent is waiting on:

a slow/hung model call
a tool call
memory/compaction
a locked/busy session from the same x-openclaw-session-key
provider timeout/fallback behavior

…the HTTP request just stays open and looks like “no response”.

I’d strongly recommend updating first — 2026.4.7 is pretty old now and a bunch of timeout/stuck-session fixes landed after that:

openclaw update

#

openclaw gateway restart

Then test with streaming so you can tell whether the run starts:

curl -N http://127.0.0.1:18789/v1/chat/completions \
  -H "Authorization: Bearer $OPENCLAW_GATEWAY_TOKEN" \
  -H "Content-Type: application/json" \
  -H "x-openclaw-session-key: agent:test-analyzer:debug-1" \
  -d '{
    "model": "openclaw/default",
    "stream": true,
    "messages": [{"role":"user","content":"Say hello briefly"}]
  }'

#

Also avoid reusing the same session key for concurrent CI requests. If request B uses the same session while request A is still running, it can queue behind A and appear stuck.

If it still hangs after updating, run:

openclaw status --all
openclaw logs --follow

and look for the run’s model/tool step near the time of the Postman request. Docs: https://docs.openclaw.ai/gateway/openai-http-api

#

Use /hooks/agent with an explicit Discord delivery target.

curl -X POST http://127.0.0.1:18789/hooks/agent \
  -H "Authorization: Bearer $OPENCLAW_HOOK_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "agentId": "my-agentId",
    "name": "CI test report",
    "message": "Analyze the test report and summarize the result.",
    "sessionKey": "hook:ci:build-123",
    "deliver": true,
    "channel": "discord",
    "to": "channel:<discord-channel-or-thread-id>"
  }'

#When making a POST request to /hooks/agent, the chat history for the same session will be cleared.