OpenViking Experience Memory
Use prior task Experience as advisory operational guidance. Keep the current
user request, current environment, and verified tool results authoritative.
Experience retrieval supplements normal context retrieval; it does not replace
user memory, events, preferences, session archives, resources, or Agent Skills.
Preconditions
- Use only OpenViking tools that are actually registered in the current Agent
runtime.
- Require both a semantic search capability and an exact URI read capability.
- If either capability is unavailable, continue without Experience. Do not
invent a tool call, fabricate a ToolPart, or replace the Agent tool call with
direct HTTP or CLI access.
- Do not substitute broad memory
recall for a search scoped to the Experience
root. Continue using the runtime's normal recall and retrieval flows when the
task also needs user facts, events, decisions, prior conversation, or domain
resources.
Select Runtime Tools
Choose the registered names that match the current runtime:
| Runtime |
Search |
Read |
| OpenViking MCP, Codex, Claude Code |
find or search |
read |
| OpenCode |
openviking_find or openviking_search |
openviking_read |
| OpenClaw |
ov_search |
ov_read or ov_multi_read |
The host may display MCP names with a namespace such as
mcp__openviking__find. Use the exact registered name and schema shown by the
runtime. Prefer find for a fast task-start lookup and search when session
context or deeper intent analysis is useful.
Retrieval Workflow
Decide whether the request is an executable task. Retrieve Experience for
planning, tool use, environment changes, multi-step workflows, or recovery
from a failed attempt. Skip retrieval for casual conversation and simple
knowledge answers.
Build one concise query containing the task goal, domain object, intended
operation, and important constraints. After a failure, include the failed
operation and stable error signature.
Search only the current user's Experience root:
viking://~/memories/experiences
For tools using MCP-style parameters, set target_uri to this root. For
OpenClaw ov_search, set uri to this root. Never hardcode default,
test, or another user ID.
Start with limit=5 and the tool's normal score threshold. Judge results by
task, environment, preconditions, and likely effect; title similarity alone
is insufficient. If no result is relevant, continue without Experience and
do not broaden the search to unrelated memory directories.
Select only the one to three Experience files likely to change execution.
Require an exact file URI without a query or fragment. Ignore directories,
unrelated memory types, and sidecars such as .abstract.md and .overview.md.
Read every selected canonical viking://.../memories/experiences/... URI
with the runtime's OpenViking read tool. Search abstracts help selection but
are not a substitute for reading the Experience body.
Apply relevant steps and checks while executing the task. Do not repeat the
Experience verbatim to the user unless its content is directly needed in the
answer.
If execution fails for a materially new reason, perform at most one focused
follow-up search using the failure evidence, then read only newly relevant
Experience files.
Applying Retrieved Experience
- Treat Experience as reusable procedure, not as a user profile, user intent,
security policy, or proof that an action succeeded.
- Follow priority in this order: system and developer instructions, current
user request, current environment and tool evidence, then Experience.
- Ignore stale, incompatible, unsafe, or conflicting guidance. Verify commands,
paths, APIs, versions, and destructive actions against the current task.
- Preserve confirmation requirements and permission boundaries. Prior success
never authorizes a destructive or external action in the current session.
- When multiple Experience files conflict, prefer the one whose preconditions
match the current environment; otherwise proceed conservatively and surface
the ambiguity when it affects the user.
Session Evidence
Use real Agent tool calls so the committed OpenViking session retains their
ToolParts:
- A completed generic OpenViking
find, search, or list result containing
an Experience URI records recall for that Experience.
- A completed generic OpenViking
read or multi_read of an Experience URI
records injection and can associate the resulting trajectory with that
Experience.
- Failed, cancelled, or incomplete calls do not count.
Do not edit, summarize away, or synthesize these ToolParts before the session
is committed.
Example
For a request to fix a deployment failure:
- Search the Experience root with a query such as
Kubernetes deployment image pull failure private registry.
- Read the most relevant exact Experience URI.
- Check that its registry, credential, and rollout assumptions match the
current cluster.
- Apply the compatible diagnostic steps, verify the live result, and continue
the user's task.
1---2name: ov-experience-memory3description: Retrieve and apply OpenViking Experience memories through the Agent runtime's generic OpenViking search and read tools. Use before or during executable, multi-step, or tool-based work such as coding, file or data changes, configuration, deployment, workflow execution, and failure recovery when prior operational guidance could improve reliability. Do not use for casual chat or simple factual questions.4---5
6# OpenViking Experience Memory
7
8Use prior task Experience as advisory operational guidance. Keep the current
9user request, current environment, and verified tool results authoritative.
10Experience retrieval supplements normal context retrieval; it does not replace
11user memory, events, preferences, session archives, resources, or Agent Skills.
12
13## Preconditions
14
15- Use only OpenViking tools that are actually registered in the current Agent
16 runtime.
17- Require both a semantic search capability and an exact URI read capability.
18- If either capability is unavailable, continue without Experience. Do not
19 invent a tool call, fabricate a ToolPart, or replace the Agent tool call with
20 direct HTTP or CLI access.
21- Do not substitute broad memory `recall` for a search scoped to the Experience
22 root. Continue using the runtime's normal recall and retrieval flows when the
23 task also needs user facts, events, decisions, prior conversation, or domain
24 resources.
25
26## Select Runtime Tools
27
28Choose the registered names that match the current runtime:
29
30| Runtime | Search | Read |
31| --- | --- | --- |
32| OpenViking MCP, Codex, Claude Code | `find` or `search` | `read` |
33| OpenCode | `openviking_find` or `openviking_search` | `openviking_read` |
34| OpenClaw | `ov_search` | `ov_read` or `ov_multi_read` |
35
36The host may display MCP names with a namespace such as
37`mcp__openviking__find`. Use the exact registered name and schema shown by the
38runtime. Prefer `find` for a fast task-start lookup and `search` when session
39context or deeper intent analysis is useful.
40
41## Retrieval Workflow
42
431. Decide whether the request is an executable task. Retrieve Experience for
44 planning, tool use, environment changes, multi-step workflows, or recovery
45 from a failed attempt. Skip retrieval for casual conversation and simple
46 knowledge answers.
472. Build one concise query containing the task goal, domain object, intended
48 operation, and important constraints. After a failure, include the failed
49 operation and stable error signature.
503. Search only the current user's Experience root:
51
52 ```text
53 viking://~/memories/experiences
54 ```
55
56 For tools using MCP-style parameters, set `target_uri` to this root. For
57 OpenClaw `ov_search`, set `uri` to this root. Never hardcode `default`,
58 `test`, or another user ID.
594. Start with `limit=5` and the tool's normal score threshold. Judge results by
60 task, environment, preconditions, and likely effect; title similarity alone
61 is insufficient. If no result is relevant, continue without Experience and
62 do not broaden the search to unrelated memory directories.
635. Select only the one to three Experience files likely to change execution.
64 Require an exact file URI without a query or fragment. Ignore directories,
65 unrelated memory types, and sidecars such as `.abstract.md` and `.overview.md`.
666. Read every selected canonical `viking://.../memories/experiences/...` URI
67 with the runtime's OpenViking read tool. Search abstracts help selection but
68 are not a substitute for reading the Experience body.
697. Apply relevant steps and checks while executing the task. Do not repeat the
70 Experience verbatim to the user unless its content is directly needed in the
71 answer.
728. If execution fails for a materially new reason, perform at most one focused
73 follow-up search using the failure evidence, then read only newly relevant
74 Experience files.
75
76## Applying Retrieved Experience
77
78- Treat Experience as reusable procedure, not as a user profile, user intent,
79 security policy, or proof that an action succeeded.
80- Follow priority in this order: system and developer instructions, current
81 user request, current environment and tool evidence, then Experience.
82- Ignore stale, incompatible, unsafe, or conflicting guidance. Verify commands,
83 paths, APIs, versions, and destructive actions against the current task.
84- Preserve confirmation requirements and permission boundaries. Prior success
85 never authorizes a destructive or external action in the current session.
86- When multiple Experience files conflict, prefer the one whose preconditions
87 match the current environment; otherwise proceed conservatively and surface
88 the ambiguity when it affects the user.
89
90## Session Evidence
91
92Use real Agent tool calls so the committed OpenViking session retains their
93ToolParts:
94
95- A completed generic OpenViking `find`, `search`, or `list` result containing
96 an Experience URI records recall for that Experience.
97- A completed generic OpenViking `read` or `multi_read` of an Experience URI
98 records injection and can associate the resulting trajectory with that
99 Experience.
100- Failed, cancelled, or incomplete calls do not count.
101
102Do not edit, summarize away, or synthesize these ToolParts before the session
103is committed.
104
105## Example
106
107For a request to fix a deployment failure:
108
1091. Search the Experience root with a query such as
110 `Kubernetes deployment image pull failure private registry`.
1112. Read the most relevant exact Experience URI.
1123. Check that its registry, credential, and rollout assumptions match the
113 current cluster.
1144. Apply the compatible diagnostic steps, verify the live result, and continue
115 the user's task.
gabrielmoreira